A moderating effect of Facebook time and Facebook friends on the relationship between political marketing factors and voters' behaviour in Malaysia

Transcription

1 Southern Cross University Theses 2018 A moderating effect of Facebook time and Facebook friends on the relationship between political marketing factors and voters' behaviour in Malaysia Bin Abd Latif Mohd Radzi Southern Cross University Publication details Mohd Radzi, BAL 2018, 'A moderating effect of Facebook time and Facebook friends on the relationship between political marketing factors and voters' behaviour in Malaysia', DBA thesis, Southern Cross University, Lismore, NSW. Copyright BAL Mohd Radzi 2018 epublications@scu is an electronic repository administered by Southern Cross University Library. Its goal is to capture and preserve the intellectual output of Southern Cross University authors and researchers, and to increase visibility and impact through open access to researchers around the world. For further information please contact epubs@scu.edu.au.

2 A Moderating Effect of Facebook Time and Facebook Friends on the Relationship between Political Marketing Factors and Voters Behaviour in Malaysia MOHD RADZI BIN ABD LATIF A research thesis submitted to the Graduate College of Management, Southern Cross University, Australia, in partial fulfilment Of the requirements for the degree of Doctor of Business Administration Date: 18 th May 2018

3 Statement of Original Authorship I certify that the substance of this thesis has not currently been submitted for any degree and has not previously been submitted for any other degree. I acknowledge that I have read and understood the University s rules, requirements, procedures and policy relating to my thesis. I also certify that to the best of my knowledge any help received in preparing this thesis and all sources used have been acknowledged in this thesis. Mohd Radzi Bin Abd Latif Date: 18 th May 2018 II

4 ABSTRACT Facebook and other popular social media have had a substantial impact on political marketing in developed countries. In Malaysia, Facebook was providing tools and data that could influence and track voters in their decision to identify with parties and select candidates. As a result, all parties and candidates are incorporating Facebook and other social media into their campaign strategies. But what role was social media playing in political marketing and how was Facebook actually impacting voter decisions in Malaysia? The objective of this study is to extrapolate the outcome of the next general election in Malaysia by establishing a political marketing model that can predict how political marketing strategies can influence Facebook fans and determine other factors that impact their voting. Exchange Theory was used to test the influence of political marketing factors on voter decisions, and political activities were hypothesised as exogenous variables and voting decisions as an endogenous variable. A quantitative methodology was employed. A total of 505 respondents who voted in 2008 election were selected using random sampling technique. The hypothesised model was validated using a random sample of 150 respondents from the most recent election (2013). Structural Equation Modeling (SEM) technique using Analysis of Moment Variance (AMOS) succeeded in establishing and validating an empirical model of voting decisions in Malaysia. 1 st order CFA, Confirmatory Factor Analysis, hypothesised model, generated model and also re-specified model analysis succeeded in validating the detailed measurements. Moderated Structural Equation Modeling (MSEM) using Partial Least Square (PLS) technique with the application SMART PLS also succeeded in testing the moderating effect of Facebook Time and Facebook Friends on the relationship between political marketing factors and voter decision. Interaction effect analysis between predictors and the moderators succeeded in explaining voter decision as an endogenous variable. Path analysis of a re-specified model confirmed a direct positive significant influence of Internet utilisation, candidate profiles, III

5 political aptitude, party orientation and political activities on voters decision. Analysis of political marketing factors achieved research objectives by explaining voter decisions and confirming Facebook Time and Facebook Friends as moderating variables. The present study came to a conclusion that political marketing factors should be considered as a primary tools for predicting voter decisions in future elections. For future elections, social media like Facebook must be prioritised in an effort to influence the behaviour of young voters. The fundamental findings of this study had contributed to the body of knowledge that illustrate the impact of political marketing predictors on the voter decisions. For further study, using a triangulation approach is strongly recommended to investigate the balance between quantitative and qualitative methodology in political marketing research. Keywords: Political Marketing, Voter Decisions, Exchange Theory, and Facebook. IV

6 Acknowledgement It would not have been possible to write this doctoral thesis without the help and support of the kind people around me, who have earned my deepest appreciation. First of all, I would like to sincerely thank my parents who paved the way for these studies and helped to ensure a successful outcome. I owe a great debt to my supervisors, Professor Dr Juhary Haji Ali and especially Paul Weeks who never accepted anything less than my best efforts. Their support, guidance and comments were vital to my DBA studies. Their invaluable advice illuminated many complicated areas and issues, and for that, I thank them. They are not only knowledgeable advisors but also are wonderful friends. Further, I would like to thank Datuk Wan Ahmad bin Wan Omar (Wan Ahmad) Deputy Chairman of the Malaysian Election Commission for his insight on the detail voters data as well as sampling identification. Understanding from my family inspired and allowed me to concentrate on my studies and to write this thesis. Words cannot express my depth of gratitude to my wife, who encouraged and supported my studies. Additionally, I would like to thank all the staff at City University College of Science and Technology (City-U) who provided an excellent and conducive environment for academic studies and research. V

7 TABLE OF CONTENTS CHAPTER 1 INTRODUCTION Introduction Background of the Study Problem Statement Research Objectives Research Questions Significance of the Study Thesis Structure 21 CHAPTER 2 LITERATURE REVIEW Introduction Definition of Politics Definition of Marketing The Concept of Political Marketing Underpinning Theory Social Media Facebook The Concept of Facebook Facebook Friends Facebook Time Discussion Political Marketing Overview of the Malaysian Political Scene Measurements in the Research Predictors of Voters Behaviour The Internet and Political Marketing Facebook and Political Activity Candidate Profiles Political Aptitude 62 VI

8 2.6.5 Party Orientation Voter Orientation Voters Political Activity Conceptualised Relationship and Influence between Predictors on 65 Voters Behaviour The Impact of Internet Utilisation on Voters Behaviour The Impact of Candidate Profiles on Voters Behaviour The Impact of Voters Political Aptitude on Their Behaviour The Impact of Party Orientation on Voters Behaviour Voters` Political Activities towards Voters Behaviour Conceptual Development of the Political Marketing Model Summary 97 CHAPTER 3 RESEARCH METHODOLOGY Introduction Research Design Theoretical Framework Research Hypothesis Variables and Measurements Population and Samples Data Collection Techniques Questionnaire Design Sampling Mechanism Validation of Data Gathering Data Analysis Procedure Structural Equation Modelling (SEM) Goodness of Fit Indexes Composite Reliability and Average Variance Extracted (AVE) Multicollinearity Test 122 VII

9 CHAPTER 4 RESEARCH FINDINGS Introduction Profile of Respondents Structural Model Analysis Hypothesised Model Generated Model Re-Specified Model of Voters Decision Competing Model using Moderating Variables Moderating Effect of Facebook Time Moderating Effect of Facebook Friends Results of Hypothesis Testing Achievement of Research Objectives Model Validation using Current Election Data (GE13) CFA of Exogenous Variables CFA of Endogenous Variables Summary 177 CHAPTER 5 VALIDATION AND RELIABILITY TESTS Data Screening Multivariate Outliers Normality Test of Measurements Pilot Test Results Reliability of Internet Utilisation Reliability of Candidate Profiles Reliability of Party Orientation Reliability of Political Aptitude Reliability of Political Activity Reliability of Voters Decision Reliability Test Results of Actual Data GE Reliability of Internet Utilisation Reliability of Candidate Profiles 195 VIII

10 5.3.3 Reliability of Party Orientation Reliability of Political Aptitude Reliability of Political Activity Reliability of Voters Decisions Confirmatory Factor Analysis (CFA) of Variables (GE12) CFA of Exogenous Variables CFA of Endogenous Variables Measurement Model of Exogenous Variables Goodness of Fit Indexes (GFI), Composite Reliability (CR) and 224 Average Variance Extracted (AVE) as Discriminant Validity Testing 5.7 Model Validation using Current Election Data GE CFA of Measurements CFA of Exogenous Variables CFA of Endogenous Variables Summary 245 CHAPTER 6 DISCUSSION OF THE FINDINGS The Context of this Research Introduction Theoretical Development of the Political Marketing Model Political Marketing Exchange Theory Social Marketing Usage & Political Marketing in Malaysia Political Marketing Development in Malaysia Structure of Political Marketing Model Analysis Variables and Measurements Demographic and Technographic Analysis Moderating Effect of Facebook Time and Facebook Friends Empirical Analysis on Political Marketing Development in 261 Malaysia Outcome and Data Modeling Summary 267 IX

11 CHAPTER 7 CONCLUSIONS AND RECOMMENDATIONS Introduction Findings Regarding Social Media Usage and Political Marketing 269 in Malaysia 7.3 Conclusion Recommendations Recommendations Related to the Body of Knowledge Recommendations to the Practitioners Contribution to the Study Contribution to the Body of Knowledge Limitations of the Study Suggestion for Further Research Summary 286 REFERENCES 287 APPENDIXES 308 Appendix 1: Questionnaires 308 Appendix 2: Reliability Test of Pre-Test and Pilot Test using SPSS 328 Appendix 3. Reliability of Actual Data (N=505) 336 Appendix 4. Reliability Test for 2013 Election 344 Appendix 5: CFA Test using AMOS 352 Appendix 6: Re-Specified Model using AMOS 394 X

12 LIST OF TABLES 1.1 A Comparison of General Election Results and Facebook Usage in Malaysia 1.2 Synchronisation of the Study (Linked between Research Objectives, Research Questions and Hypothesis) Malaysia s Top 5 Social Media Channels in Model Fit Indexes Identification Source Profiles of Respondents Standardised Regression Weights of Re-Specified Model Factor Loading of Re-Specified Model Summary of Moderating Effect Analysis Summary of Hypothesis Testing Standardised Regression Weights of Internet Utilisation Standardised Regression Weights of Candidate Profiles Standardised Regression Weights of Party Orientation Standardised Regression Weights of Political Aptitude Standardised Regression Weights of Voters Like Standardised Regression Weights of Voters Decision a Normality Test and Identification of Non-Normal of Measurement b Final Transformation of Non-Normal Items a Reliability Statistics of Internet Utilisation b Item-Total Statistics of Internet Utilisation 184 XI

13 5.3a Reliability Statistics of Candidate Profiles b Item-Total Statistics of Candidate Profiles a Reliability Statistics of Party Orientation b Item-Total Statistics of Party Orientation a Reliability Statistics of Product Orientation b Item-Total Statistics of Product Orientation a Reliability Statistics of Sales Orientation b Item-Total Statistics of Sales Orientation a Reliability Statistics of Marketing Orientation b Item-Total Statistics of Marketing Orientation a Reliability Statistics of Political Aptitude b Item-Total Statistics of Political Aptitude a Reliability Statistics of Political Activity b Item-Total Statistics of Political Activity a Reliability Statistics of Voters Decision b Item-Total Statistics of Voters Decision a Reliability Statistics of Internet Utilisation b Item-Total Statistics of Internet Utilisation a Reliability Statistics of Candidate Profiles b Item-Total Statistics of Candidate Profiles a Reliability Statistics of Party Orientation b Item-Total Statistics of Party Orientation a Reliability Statistics of Product Orientation 198 XII

14 5.14b Item-Total Statistics of Product Orientation a Reliability Statistics of Sales Orientation b Item-Total Statistics of Sales Orientation a Reliability Statistics of Marketing Orientation b Item-Total Statistics of Marketing Orientation a Reliability Statistics of Political Aptitude b Item-Total Statistics of Political Aptitude a Reliability Statistics of Political Activity b Item-Total Statistics of Political Activity a Reliability Statistics of Voters Decision b Item-Total Statistics of Voters Decision Standardised Regression Weights of Internet Utilisation Standardised Regression Weights of Candidate Profiles Standardised Regression Weights of Party Orientation Standardised Regression Weights of Political Aptitude Standardised Regression Weights of Voters Like Standardised Regression Weights of Voters Decision Composite Reliability and Average Variance Extracted from Measurements Composite Reliability and Variance Extracted of Variables Standardised Regression Weights of Internet Utilisation Standardised Regression Weights of Candidate Profiles Standardised Regression Weights of Party Orientation (2nd Order CFA) 233 XIII

15 5.31 Standardised Regression Weights of Political Aptitude Standardised Regression Weights of Voters Like Standardised Regression Weight of Voters Decision 242 XIV

16 LIST OF FIGURES 1.1 A Comparison of Business Marketing and Political Marketing The Link between Problem Statement and Research Questions Basic Conceptual Model Scope of the Study: Exchange Theory in Marketing & Politics Review Flow and Conceptual Development of the Hypothesised Model Political Marketing Strategies and Votes Getting a Political Marketing Strategies and Facebook Fans b Research Journey on the Voters Behaviour Model Framework for Political Parties Developed Countries Marketing Communications Framework for Small Political Parties Individual Political Choice Behaviour Ideal Political Marketing Orientation Profile Marketing Communications Framework for Small Political Parties Strategic Marketing Communications Framework for Small Political Parties Business Marketing and Political Marketing Compared Model of Political Planning for Local Campaigning Theory of Planned Behaviour Relationship Marketing in Predictive Models in the Political Process 87 XV

17 2.15 Channels of Party Communication Political Marketing Process Buyer Decision Processes Candidate Marketing Map Newman s Model of Political Marketing Model of Political Communication The Alternative Model of Political Marketing Strategic Process for Segmenting Political Markets: the STP Approach Political Marketing Model Navigation System and Functionality of the Reserved Area of the Site Evolution of the Brand Construction Process a Research Design and Model Journey b Hypothesised Model c Hypothesised Model with Moderating Variables SEM Procedure Flowchart of Covariance of the Structural Model Hypothesised Model using SEM Hypothesised Model between Exogenous and Endogenous Variables a Generated Model of Voters Decision b Re-Specified Model of Voters Decision Moderating Effect of Facebook Time 142 XVI

18 4.5 Moderating Effect of Facebook Friends a CFA of Internet Utilisation b CFA of Internet Utilisation CFA of Candidate Profiles a CFA of Party Orientation b CFA of Party Orientation for GE a CFA of Political Aptitude b CFA of Political Aptitude a CFA of Political Activities b CFA of Political Activities a CFA of Voters Decision b CFA of Voters Decision a CFA of Internet Utilisation b CFA of Internet Utilisation a CFA of Candidate Profiles b CFA of Candidate Profiles a CFA of Party Orientation b CFA of Party Orientation a CFA of Political Aptitude b CFA of Political Aptitude a CFA of Political Activities b CFA of Political Activities a CFA of Voters Decision 221 XVII

19 5.6b CFA of Voters Decision Measurement Model of Exogenous Variables a CFA of Internet Utilisation b CFA of Internet Utilisation CFA of Candidate Profiles a 5.10b CFA of Party Orientation CFA of Party Orientation For GE a CFA of Political Aptitude b CFA of Political Aptitude a CFA of Political Activities b CFA of Political Activities a CFA of Voters Decision b CFA of Voters Decision Voters Decision Model Facebook Political Marketing Model (FPMM) 281 XVIII

20 LIST OF ABBREVIATIONS AGFI AIC AVE CFA CFI CMIN C.R DF EFA GFI GOF KMO MD M.I ML NFI RMSEA S.E SEM Skew SMC TLI VE Adjusted Goodness of Fit Index Aikake Information Criterion Average Variance Extracted Confirmatory Factor Analysis Comparative Fit Index Minimum Discrepancy Critical Ratio Degrees of Freedom Exploratory Factor Analysis Goodness of Fit Index Goodness of Fit Kaiser-Meyer-Olkin Mahalanobis Distance Modification Indices Maximum Likelihood Normed Fit Index Root Mean Square Error of Approximation Standard Error Structural Equation Modeling Skewness Squared Multiple Correlations Tucker Lewis Index Variance Extracted XIX

21 CHAPTER 1 INTRODUCTION 1.1 Introduction One recent approach that political candidates are adapting, to be competitive, is to attract voters through social media like Facebook. They first create a page, which is mostly devoted to running campaign activities. They try to design their pages based on voters interests, and then they start establishing rapport with voters. Voters now have the opportunity to interact with a candidate as a member of a group or on a one-to-one basis. They can state their demands and ask candidates to present an overview of their intentions. Through discovering voters preferences, not only do candidates have the opportunity to motivate their voters to vote for them, but they also seek to gain a better indication of the probability of their success. After the 2008 Malaysian 12 th general election (GE12), the former Prime Minister, Abdullah Ahmad Badawi has acknowledged the power of social media and admitted after the 2008 Malaysian 12 th general election (GE12), that Barisan Nasional (BN), the government coalition, had suffered defeat to the opposition in the online war, possibly because of BN underestimating the power of social media. (Barisan Nasional is a multi-racial coalition, with the major players being the Malaysian Indian Congress (MIC), Parti Gerakan Rakyat Malaysia (GERAKAN), the Malaysian Chinese Association (MCA) and United Malay National Organisation (UMNO) as the major coalition partner). Badawi also stated that, in 2008, as in the earlier elections, BN had been prioritising the traditional media and had disregarded the importance of social media that had gradually witnessed BN s loss of 2/3 majority in the Parliament (Sani, 2014). The 2/3 majority constitutes a vital fraction, as this is the majority required to be able to amend the Constitution 1

22 as required in the article 159(3) of the Malaysian Constitution. After the 12th election in 2008, both government and the opposition have confirmed the importance and the effect of social media in Malaysia. Both parties are active participants in Facebook, Twitter, and blogs. Leaders popularity is determined by how many likes or followers they have on social media. It has come to the point when social media is seen as reshaping the Malaysian political landscape. Prime Minister Najib Tun Abdul Razak was quoted as saying that the 13th general election could be regarded as Malaysia s first social media centric election (Lim, 2013). Our current prime minister is more social media savvy than his predecessors, as he is a keen user of Facebook and Twitter. Samad (2013) conceded that Prime Minister Najib had 1,310,676 followers on Twitter, and his Facebook page had 1,366,187 likes. He constantly uploaded postings of his photos and events in Furthermore, his 1Malaysia blog also has many followers, and was constantly updated. This study focuses on political marketing and its principal applications in Malaysian politics through social media, particularly Facebook. It examines the voting considerations and behaviour of young voters between the ages of 21 to 30 years, who are the more computersavvy group in Malaysian society. Tang (2009) conceded that more than 60 % of those between 21 to 40 years followed blogs and online media for reliable information as opposed to only 23 % who regarded the information from television programs are more reliable, and % share the same thought about newspapers. This study attempts to identify the factors that actually determine voting decisions of Facebook fans, as representatives of potential mainstream voters in Malaysian general elections. Possible factors are numerous; hence, for the purpose of this study, the scope has been narrowed down to those factors related to Political Marketing Exchange Theory 2

23 henceforth referred to as PMET. The study questions whether PMET works in Malaysia and if it does not, what are the major factors taken into account by young voters in Malaysia when voting. The study also considers whether it is possible to predict Facebook users voting behaviour as voters for the next general election. 1.2 Background of the Study Preparing for the 2013 election (GE13), BN, the government coalition and the opposition coalition of three parties with the members being the Pakatan Rakyat (PR) People Justice Party (PKR), Pan-Malaysian Islamic Party (PAS) and The Democratic Action Party (DAP) consciously assumed that cyberspace would be the critical battleground in the polls. It was shown by BN s own intelligence that more than 60 per cent of the voters under 35 were undecided voters and Pakatan parties had also stressed the use of the Internet. Linnarz (2010) made a point that even the usually serious PAS uploaded three videos to YouTube in February 2012 in preparing for the elections in Kelantan, its main political battlefield. Both sides were aiming at these digital natives, with heavy dependence on the online toolbox. Despite the challenge of Bersih 2.0 in July 2011, a public rally and demonstration seeking to promote free and fair elections, BN politicians had made a great effort in revamping and strengthening their web strategy, with the younger voters in mind in order to channel their strong presence on Facebook and Twitter into physical interaction. This new media engagement is giving rise to a more informed, participatory electorate, i.e. toward democracy, even if it is unlikely that a dramatic political change will happen via any vector in Malaysia without regime change (Ormrod, 2011). The 2013 general election results published by the Malaysian Election Commission (MEC) showed that BN had lost the popular vote despite retaining a majority of parliamentary seats. 3

24 There is a statistical highlight establishing that BN coalition secured 133 seats with 47.38% of the popular vote compared to 140 seats with 50.27% of the popular vote in The loss reflected seven seats and 2.89% of the popular vote. Also, in the 2013 election, the PR coalition won 89 seats with 50.87% of the popular vote, compared to 82 seats and 46.75% of the popular vote in The coalition managed to occupy seven more seats and an increase of 4.12% of the popular vote. For parliamentary seats, BN showed its strength in nine out of thirteen states in Malaysia. The poor election performance of BN correlates with exponential growth of social media users in Malaysia over the preceding five years. During the previous election in 2008, there were 800,000 Facebook and 3,429 Twitter users in Malaysia. However, by 2013 these numbers had increased to 13,220,000 for Facebook and 2,000,000 for Twitter users (Forestinteractive.com, 2013). In terms of party level performance, GE13 changed some established patterns. BN s reduced number of total seats in GE13 reflects poorer performance by the coalition partners. UMNO, as the largest and dominant partner in the BN coalition, saw its support increase slightly, most likely due to the majority support of Malay and Bumiputera voters in the states of Sabah and Sarawak. Hence BN s continued existence in peninsular Malaysia depended on UMNO, which survives on the Malay voters, while at the same time, the coalition was gradually losing its Chinese and Indian voters. GE13 even saw the defeat of a number of top leaders of BN, including two long-serving chief ministers of the states of Melaka and Johor. The chief minister of Johor made a special request to the voters of his constituency through a full-page newspaper advertisement. However, despite this, as well as continuous attacks on 4

25 the opposition DAP by MCA through dozens of full-page newspaper advertisements, several of the long established BN leaders lost their seats. This could reflect the significance of the young voters and the change of mindset about leadership. BN s recruitment strategy of selecting new and young candidates for one-third of the seats have possibly prevented further loss of a number of other parliamentary seats. The performance of the opposition alliance, PR, was much weaker than expected even though they gained an additional seven parliamentary seats. The opposition had touted GE13 as the Road to Putrajaya ; with growing confidence that voters overwhelmingly wanted a change in the government and that they would definitely win a majority of seats and take over as the federal government. However, the election results frustrated their expectations. The candidacy and voting patterns suggest that the parties operated basically along ethnic lines, and the voters also largely remained along the same ethnic lines. Since BN members contested under the banner of BN instead of separate party banners, the voters went for the candidates chosen by the BN, although ethnicity still played a major role. Predictably, no independent candidate won any seat signifying the irrelevance of non-party electoral contests. The results of GE13 caused the major parties in the BN coalition to re-calculate their expectations, threats and opportunities. Despite substantial increase in subsidies, economic and social welfare, hard cash assistance, dozens of national economic development programs that the BN government had introduced during the past four years of Najib s rule, the ruling BN coalition witnessed a further erosion of its popularity among the voters. 5

26 The coalition in GE13 performed poorly compared to the 2008 s polls. BN s two major coalition partners, MCA and GERAKAN, suffered miserably. The two elections made it clear that the political culture of Malaysia was shifting to a more a participatory type with an increasing rate of voters turnout at 84.5 per cent and a younger generation of voters with increased social and economic mobility. With the relaxation of some political restrictions such as greater freedom of expression made possible by censorship-free social media platforms, this participatory trend is likely to strengthen in future. These fundamental changes in Malaysian politics and voting patterns form the conceptual context for this study which examines the relationship between traditional marketing, social media marketing and political marketing. While at first glance, marketing in a business setting, can be perceived as strategised activities to help improve businesses organisations, it does more than that. Recent developments suggest a transition from traditional marketing to social media marketing. Ernestad and Henriksson (2010) explained that the move towards social media marketing still continues with changes in organisations mindsets and that this transition can be more complex for larger organisations. Objectives in social media do not seem to be specified to any great extent; instead, they are more rhetorical and idealistic. Social media increases people s perception of brands available in the market, regardless of the kind of market. Similarly, in a political setting, voters can study candidates through their profiles on social media. Furthermore, they can make demands on what they need (or want), familiarise themselves with the candidates` agendas and communicate interactively with candidates who may be too busy to meet in person. Among all the social media, one of the most effective platforms to accomplish such interaction is Facebook, through friending the politicians via their Facebook pages or joining Facebook groups set up by the politicians. 6

27 The idea of mixing politics with marketing has resulted in the concept of political marketing. It has attracted much attention because it is considered to be very instrumental in contributing to the effectiveness of political activities and achieving political objectives. The basic aspect of politics is an election of candidates and voting. Candidates represent goods and/or services to the voters who are regarded as the customers. Political marketing resorts to the same techniques as other forms of marketing to introduce a concept or an idea, instead of a specific product or service, and encourage people to go for, or choose that idea (Ahmed, Lodhi, & Shahzah, 2011). Aided by political marketing, ideas are directly linked with the voters to fulfil their political needs and to offer support for the candidates and the ideas pitched (Shama, 1975). There are similarities between the marketing of goods and services and political marketing in that both rely on the notion of promotion. The considerable impact brought about by social media can be seen (Niffenegger, 1989). Through the processes of informing, reminding and changing attitudes of the customers, who are treated as voters in this context, their behaviour could be affected and in some cases manipulated. Both marketings of goods and services and marketing of political candidates utilise similar tools such as market research as well as several statistical and computing techniques adopted to study the market (Henneberg & O Shaughnessy, 2007). There are also other similarities between the two marketing practices. Concepts such as consumer behaviour, market segmentation, image and brand loyalty are some of the examples. Not only are they similar in concepts but they also utilise tools in political and business marketing. In political marketing, some tools like market research, media, and advertising are also used to evaluate and analyse the candidates (Butler, Collins, & Fellenz, 2007). 7

28 There is proof that social media is an efficient means of conveying messages to receivers and this includes political messages. It can be argued that the internet, in general, and social media in particular, has played a significant role in the realm of political marketing. Even though the initial purpose of creating such an environment is to facilitate personal interaction, it also has been effective in other contexts like a business and advertising (Hesseldahl, MacMillan, & Kharif, 2008). Due to the widespread adoption of social media, politicians, like businesspeople, decided to exploit this technology to communicate, and consequently influence those followers from whom they are seeking support. Active voters tend to develop political aptitudes, which enhance their political participation. One way to increase the level of political aptitudes is to read and follow the news, especially on the Internet through social media, such as Facebook. As Vitak et al. (2010) explained, Facebook has equipped users with the platform to practise civic skills or political activity where activities, like taking part in a political group or sharing a link, can be done with only a few mouse clicks. Hence voters no longer are limited to one-sided sources of information. Through the rapid growth in social media usage, as well as increased levels of education and insight, voters decision-making process are very much influenced by political issues (Alexey, 2006). As far as the political activities of voters are concerned, voters now use social media to participate actively in politically related issues. Social media provides users with the opportunity to express themselves and share their knowledge and opinions, and Facebook provides such an access point. Westling (2007) believed that Facebook could be put to good use to organise political agendas by sending messages on meetings or broadcasting the campaign updates, or creating a contact list of voluntary supporters among all users. Facebookers can also be invited by their friends or followers to join a group or to forward 8

29 campaign messages to their respective friends and acquaintances. According to Yousif and Alsamydai (2012), political persuasion is one of the tools candidates can use to influence voters behaviour through social media. They can also make use of Facebook to send advertisements to voters and to involve voters in campaign activities, once candidates have targeted the right voters and effective advertisements have been sent to and accepted by voters. Niffenegger (1989) suggested that modern political marketing manipulates publicity and advertising techniques, whilst simultaneously integrating the 4Ps which are the mainstay of marketing: Price, Product, Promotion and Place. Lees-Marshment (2001) noted that the three components of sales and product oriented marketing strategies could find their place equally well in politics and could become the fundamentals of political marketing. According to Lees-Marshment (2001), political marketing referred to the political organisations that practise business marketing concepts and techniques to help them materialise their goals. 1.3 Problem Statement Globally, the impact of social media, including Facebook, has interested several researchers. Robertson, Vatrapu, and Medina (2010) found that Obama, who eventually won the US presidency, received the highest number of comments on his Facebook page compared to the number received by Clinton and McCain. In the case of Australia, Macnamara (2008) found that Kevin Rudd, the winner of the 2007 Australian Federal election,had gained more public comments on his Facebook profile than his opponents. Johnson and Perlmutter (2010) concluded that the significant role of Facebook in effective political marketing is best described by the term Facebook Election. Williams and Gulati (2007, 2008, 2009), whose research tracked the relationship between Facebook activity and outcomes in US elections, including the 2006 midterm congressional elections, 2008 early nomination contest, and

30 US presidential elections, suggested that a politician s Facebook activity and support impacts significantly on his/her vote share. Effing, Van Hillegersberg, and Huibers (2011) discovered a significant relationship between a politician s involvement in social media and his vote share, using the results of the 2010 national election for the House of Representatives in the Netherlands. With specific reference to young voters, Towner and Dulio (2011) found that young voters in the 2008 U.S. presidential election were more likely to vote if they had visited the candidate s Facebook profile. Lappas et al. (2010), having investigated Facebook activity of the two Greek parties, PASOK and Nea Dimokratia, during the 2009 national election, concluded that election outcomes may have been predicted by the number of Facebook likes PASOK received compared to Nea Dimokratia, thus reflecting the inclinations of Greek voters. Leskinen (2012) examined the number of likes received by candidates during Parliamentary elections in Finland and found that the winning party (Finns Party) gained the highest number of likes compared to the other parliamentary parties. It would be interesting to establish whether a similar trend had taken place in developing countries like Malaysia. Realising the rapid growth of Internet penetration and social media usage, it can be speculated that the incremental inclusion of Facebook in political marketing had impacted traditional norms of using (Political Marketing Exchange Theory) PMET to win votes. 10

31 Table 1.1 Comparison on the General Election Results and Facebook Usage in Malaysia Year Barisan Nasional (BN) Pakatan Rakyat (PR) Facebook Users ** Negligible , ,322,000 Source for the Malaysian General Election results: Malaysian Election Commission (2013) ** Note : Facebook was launched on 4 th February 2004 and the Malaysian 11 th General Election (GE11) was held on 21 st March 2004 ; During the infancy stage of Facebook, statistics on the number of Facebook users in Malaysia were not available online. Thus the number of Facebook users is assumed to be negligible having been launched 45 days before the GE11. In Malaysia, the 2004 general election result (GE11) was considered the best ever for the ruling government in terms of the number of parliament seats gained compared to the previous general elections, with the Barisan Nasional having captured more than 2/3 majority of the parliamentary seats (Table 1.1). The 2008 election result was on the other hand, was the worst ever in terms of number of parliamentary seats lost compared to the previous elections. Both elections were under the same Prime Minister Abdullah Badawi with the same coalition component of political parties. The question that emerges is Could the drastic change in the result be explained by rapid growth in the number of Facebook users? Before the 2008 election, Malaysians practised communal politics along racial lines where the basis of voters selection was their respective communal party s credibility, promises and promotion. The 2008 election witnessed a drastic change, moving towards the candidates individual background and credibility and also a broader more abstract consideration. The issue is whether this was influenced by the extensive application of social media in the campaign and promotion by candidates who reached out to their voters, directly bypassing their communal parties traditional face-to-face communication channel and by the insurgence of the young voters who leveraged on the viral effect of social media to influence voters. 11

32 The 12th General Election in 2008 was considered an important milestone in Malaysian politics as it heralded the dawn of a new political scenario (Jawan & Yussoff, 2009). It was described metaphorically as a political volcanic eruption. It witnessed the rise of the peoples power against suppression or dissatisfactions over communal politics and deteriorating social conditions (increased cost of living, crime, and corruption) (Lim & Wai-Mun, 2008). The most dissatisfied groups in Malaysia society appear to be the youth and the young voters. They were beginning to show their teeth during the last 12 th and 13 th election. It is expected that their role will be very important in deciding the forthcoming election results. Young voters have boycotted mainstream media accusing them of being used as brainwashing tools of the government. Social media is the choice among these young voters. The primary question for this study is, Will social media be strong enough to influence the voters behaviour and hence determine the outcome of future elections? Another question is whether Facebook fans as voters in Malaysia regard the candidate himself or herself (their background and standing) as more important than the future promises or favours they made in the past? Will the voters take into account the capability of the candidate to deliver his promises in the future when casting their votes? Previously when the opposition parties candidates won the seats, were the voters able to detect the trend rationally and switched their support? Or did they take a risk trying to change the trend and the government? As suggested by the examples above, there is a theoretical gap in Political Marketing Exchange Theory (PMET), as far as voter behaviour is concerned. Here, the interaction between voters and candidates has been fixed in the context of political marketing, based on the suggestion of Lees-Marshment (2012). Thus, an empirical study is needed to fill the gaps 12

33 mentioned. PMET emphasises the postponement of consideration for a vote given by a voter to the foreseeable future (Figure 1.1). Hence voters are taking the risk of failed consideration. Communication Goods and Services Communication Promises and Favors Sellers Buyers Candidates Voters Money Votes Information Information A. Business Marketing B. Political Marketing Figure 1.1 A Comparison of Business Marketing and Political Marketing The theory equates voters with buyers in the market while the candidate is likened to the seller. Candidates sell promises of future delivery of community services and assistance to voters, while they expect the voters to pay first in terms of votes so they can win the election (Figure 1.1). An online social media group, Facebook fans, as the potential mainstream voters in Malaysia, is known to be the most critical and vulnerable among the existing voter groups. Most of them choose to disregard politics and the voting process. This research plan to focus on the young voters groups particularly Facebook users, by using quantitative approach by administering questionnaire to gauge the potential impact of Facebook on the next election result. It is anticipated that the research will not only confirm the crucial role of Facebook and other social media but also nail down the determinant of young voters behaviour in the coming election in Malaysia. 13

34 1.4 Research Objectives Three main aspects of the research will be covered: first to examine the different voting considerations and behaviour peculiar to a specific situation among Facebook fans as voters; to be able to project the role and significance of the digital/virtual world on Facebook fans as voters in Malaysia; and to be able to predict Facebook fans behaviour as voters for the next general election (Figure 1.2; Table 1.2). Specifically, the research shall attempt: 1. To determine whether Facebook fans as voters in Malaysia expect future return/rewards for their votes, based on their experience of the last general election in 2008 in Malaysia. The research will determine whether Facebook fans as voters in Malaysia had voted for candidates that have done them favours in the past. 2. To determine whether Facebook fans as voters in Malaysia regard the tangible value of promises to be more influential than the intangible value. It is the promises and ability to honour them that must be taken into account in their voting preference 3. To determine if PMET actually worked in causing the voters to cast their votes based on the assurance of the person who had the power to deliver the consideration in the future? 4. To determine whether Facebook fans as voters in Malaysia were influenced by information on the Internet in their voting behaviour in the 2008 election. 5. To determine whether the fact that the more they were exposed to the campaign blogs and Facebook, the more likely it was that they would vote for the opposition party, meaning that they were just releasing their anger in their voting behaviour? 6. To determine if voters are irrational in their voting behaviour and hence disregard PMET considerations? The next general election, i.e. GE13 is expected to be critical for the ruling coalition because the voters have now been exposed to a broader perspective and consideration in their voting behaviour. 14

35 7. To determine whether Facebook fans as voters in Malaysia regard party image as more important than the candidates background. 8. To determine the main considerations taken when casting a vote? Are the voters more concerned simply with change and hence their voting behaviour is more influenced by other considerations than those incorporated in PMET? Basic premises in the Problem statements Facebook fans have high inclination to politics but very critical Basic premises in the Problem statements Facebook fans acceptability of political marketing strategies using facebook are high Potential support/acceptibilty of political marketing strategies using facebook by the Malaysian fans are high Figure 1.2 The link between Problem Statement and Research Questions 1.5 Research Questions Figure 1.2 and Table 1.2 depicts the relationship between problem statements and research questions. The research attempts to answer some specific research questions: 1. Do Facebook fans as voters in Malaysia expect future returns/rewards for their votes? 2. Have Facebook fans as voters voted for candidates that have given them favours in the past? 3. Do Facebook fans as voters regard the candidate s background and standing as more important than future promises or favours given in the past? 4. Do Facebook fans as voters regard the tangible value of promises as more influential than the intangible value of promises? 15

36 5. Were Facebook fans as voters influenced by information obtained from the Internet when they voted during the last election? 6. Do Facebook fans as voters regard party image as more important than the candidates backgrounds? Based on the intensive review on the gaps of literature and practices, the present study hypothesised a structural model of voting decisions (Figure 1.3 and Table 1.2). There were five predictors tested as exogenous variables to the voting decision, which, in turn, serve as the endogenous variable. Internet utilisation, candidates profiles, party orientation and political aptitude were hypothesised as predictors of voting decisions (Table 1.2). Furthermore, the present study also hypothesised Facebook Time and Facebook Friends as moderating variables on the relationship between predictors and voting behaviour (Figure 1.3). Internet Utilisation Utility Product Candidate Profiles FB Time Sales Party Orientation Voting Decision Marketing Political Aptitude FB Friend Political Activities Figure 1.3 Basic Conceptual Model 16

37 Table 1.2 Synchronisation of the Study (Linked between Research Objectives, Research Questions and Hypothesis) RQ Research Questions RO Research Objectives Hypo Hypothetical Statement Pathway: Testing Predictors of Voters Behaviour RQ4 Do Facebook fans as voters in Malaysia regard the tangible value of promises as more influential than the intangible value of promises? RO2 To determine whether Facebook fans as voters in Malaysia regard the tangible value of promises to be more influential than the intangible value. It is the promises and ability to honour them that must be taken into account in their voting consideration. Hy1a Internet Utilisation has a significant direct influence on Voting Decision RQ2 Have Facebook fans as voters in Malaysia voted for candidates that have given them favours in the past? RO1 To determine whether Facebook fans as voters in Malaysia expect future return/rewards for their votes, based on their experience of the last-but-one general election (2008) in Malaysia. The research will determine whether Facebook fans as voters in Malaysia have voted for candidates that have done them favours in the past. Hy1b Candidate Profiles have a significant direct influence on Voting Decision RQ6 Do Facebook fans as voters in Malaysia regard party image as more important than the candidates backgrounds? RO7 To determine whether Facebook fans as voters in Malaysia regard party image as more important than the candidates background. Hy1c Party Orientation has a significant direct influence on Voting Decision RQ3 Do Facebook fans as voters in Malaysia regard the candidate background and standing as more important than future promises or favours given in the past? RO6 To determine if voters are irrational in their voting behaviour and hence disregard the PMET considerations? The next general election i.e. GE13 is expected to be critical for the ruling coalition because the voters have now opened their eyes to a. Hy1d Political Aptitude has a significant direct influence on Voting Decision RQ1 Do Facebook fans as voters in Malaysia expect future returns/rewards for their votes? RO3 To determine if PMET actually worked that caused the voters to cast their votes based on the assurance of the person who had the power to deliver the consideration in the future? Hy1e Political Activities have a significant direct influence on Voting Decision 17

38 Pathway: Construction of Party Orientation Dimensions RQ6 Do Facebook fans as voters in Malaysia regard party image as more important than the candidates backgrounds? RO7 To determine whether Facebook fans as voters in Malaysia regard party image as more important than the candidates background. Hy2 Product, Sales and Marketing confirm the construct of Party Orientation Pathway: Moderating Test of Facebook Time RQ4 Do Facebook fans as voters in Malaysia regard the tangible value of promises as more influential than the intangible value of promises? RO5 To determine whether the fact that the more they were exposed to the campaign blogs and Facebook, the more likely it was that they would vote for the opposition party, meaning that they were just releasing their anger in their voting behaviour? Hy3a Hy3b Hy3c Facebook Time has a moderating influence on the relationship between Internet Utilisation and Voting Decision. Facebook Time has a moderating influence on the relationship between Candidate Profiles and Voting Decision Facebook Time has a moderating influence on the relationship between Party Orientation and Voting Decision Hy3d Facebook Time has a moderating influence on the relationship between Political Aptitude and Voting Decision Hy3e Facebook Time has a moderating influence on the relationship between Political Activities and Voting Decision Pathway: Moderating Test of Facebook Friends RQ5 Were Facebook fans as voters in Malaysia influenced by information obtained from the Internet when they voted during the last election? RO4 To determine whether Facebook fans as voters in Malaysia were influenced by information on the Internet in their voting behaviour in the 2008 Malaysian election. Hy4a Hy4b Hy4c Facebook Friends have a moderating influence on the relationship between Internet Utilisation and Voting Decision Facebook Friends have a moderating influence on the relationship between Candidate Profiles and Voting Decision Facebook Friends have a moderating influence on the relationship between Party Orientation and Voting Decision Hy4d Facebook Friends have a moderating influence on the relationship between Political Aptitude and Voting Decision Hy4e Facebook Friends have a moderating influence on the relationship between Political Activities and Voting Decision Pathway: Establishment Voters Behaviour Model RQ1 Do Facebook fans as voters in Malaysia expect future returns/rewards for their votes? RO8 To determine the main considerations taken when casting a vote? Are the voters more concerned simply with change and hence their voting behaviour is more influenced by other considerations than those incorporated in PMET? Hy5 Internet Utilisation, Candidate Profiles, Party Orientation, Political Aptitude, Political Activities, moderated by Facebook Time and Facebook Friends confirm the significant interactions in the Structural Model. 18

39 1.6 Significance of the Study The growing importance of the Internet as a political marketing tool will also shift attention towards the younger generation who are more Internet savvy than older voters. Those newly registered as voters tend to be the neutral ones and will be swayed by circumstances just before the election. In the past few elections, it has been shown that these voter groups were the prime determinants of the general election result to such an extent that all political parties in Malaysia have focused on the young voters. The result of the Malaysian 12 th general election (GE12) held in 2008 has especially, highlighted the apparent impression that Facebook fans, as voters do not even expect any short-term returns for their votes. Some of these voters were observed using their votes as a way to release their anger and frustration against certain political parties. Some were counting on the belated rewards when they voted for the Islamic party, whereas some voted for past favours or proven track record of the candidates. In the subsequent general election held in 2013 (GE13), a new group of about 2.6 million Malaysians registered to cast their ballots for the first time. This number is much higher than the 638,000 newly registered voters in GE12. These upsurge in interest in politics was triggered by the opposition s best-ever election showing in 2008 which caused more young people to register as voters. These young voters are a crucial and possibly decisive, source of support in an election swaying the votes to one coalition or the other (Sithrapuhran & Raghu, 2013). The issue at hand is whether these young voters, as Facebook fans have the PMET considerations in mind when they cast their votes. It is being anticipated that the research will discover some form of fundamental relationship between variables relating to PMET in the Malaysian political context. The present study does not expect to refute the PMET proposition on the exchange of votes with favours. However, it 19

40 may be able to refute the theory with regard to future promise and current consideration and perhaps replace it with another conceptual theory. Empirical evidence showed two obvious trends that needed to be investigated by this research. Firstly: a sudden surge in the numbers of newly registered young voters in the 12 th General Election (GE12) held in 2008 as compared to the 11 th General Election (GE11) held in Analysis of data provided by the Malaysian Election Commission revealed that the number of newly registered young voters in GE12 was slightly above 500,000, which represents around 50 per cent of 1,166,042 newly registered voters. Comparatively, in GE11, the total number of newly registered voters were merely 192,027 (Malaysian Election Commission, 2004, 2008). This surge in numbers of newly registered young voters had impacted the result of GE12, which saw the end of 2/3 rd parliamentary majority dominance that BN government had enjoyed since the first Malaysian General Election, which was held in the year Second: the accompanying popularity of Facebook in between GE11 held in 2004 and GE12 held in 2008 especially among youths. Users of Facebook in Malaysia in 2004, the year Malaysia held its 11 th General Election, were negligible since Facebook was launched that same year. However, the number of Facebook users had increased to 800,000 by the year 2008 which was the year Malaysia held its 12 th General Election. Malaysian youths were the majority of Facebook users in 2012, forming 82 per cent of Facebook users in that year. (Source: Allfacebook.com/Facebook-demographic-statistics). The empirical data deduced from the two trends reinforce the significance of this research so as to discover the fundamental relationship between variables relating to PMET in the Malaysian political context. The outcome of this research could reveal the relationship of 20

41 variables affecting young voters behaviour caused by the increasing popularity of Facebook among youths in Malaysia and the surge of young voters political activism as demonstrated by the huge number of young voters registering themselves as voters. The end of 2/3 rd majority dominance of the ruling Barisan Nasional (BN) party in Parliament, following the result of Malaysia s 12 th General Election in 2008, impacted Malaysia in several ways, including triggering the resignation of Abdullah Badawi, the 5th Malaysian Prime Minister, who was replaced by Najib Razak (Najib). To practitioners in politics, the result of this study is intended to serve as a framework for political marketing effort using Facebook and other social media channel in the future. The model to be generated by this study is intended to contribute towards the body of knowledge in the area of Facebook usage in political marketing. 1.7 Thesis Structure The present study presented in six chapters. Chapter 1 is the Introduction, outlining the research. Chapter 2, is a Literature Review that explores the current state of the literature in related areas and outlines the development of the proposed research and hypotheses. Chapter 3 is the Research Methodology chapter, which discusses the research model and relates it to the hypotheses and objectives of the study. It describes in some detail the research methodology applied. Chapter 4 contains Findings and Data Analysis. It presents the initial findings in the form of data tabulation and implications. The present study succeeds in establishing a structural model 21

42 of voters behaviour using 2008 election data (GE12) and also validates the findings with a series of 2013 election (GE13) data. Advanced quantitative data analysis and the resulting implications are also dealt with in this chapter. Chapter 5 concerns the process of testing and touches on the validation of the structural model. A series of reliability and validity tests are conducted for the confirmation of significant items of measurements and variables according to the hypotheses. Chapter 6 presents and discusses the findings and their implications. Chapter 7 deliberates on the conclusions and recommendations for future research 22

43 CHAPTER 2 LITERATURE REVIEW 2.1 Introduction In the literature review, the focus is on literature related to political marketing strategies and Facebook. The discussion begins with Exchange Theory in marketing as the underpinning theory and relates it to political marketing and how the relevant issues can be developed (Figure 2.1). The review then focuses on the literature that relates to Facebook as an important platform that may influence fans to translate their personal relationship with the candidates, and familiarity with the parties into votes in the election. Marketing Political Marketing Politics Political Marketing strategies Social Media Voters Facebook Figure 2.1 Scope of the Study: Exchange Theory in Marketing & Politics The chapter introduces the concept of politics, marketing and political marketing. It then discusses the underpinning theories and their application in political marketing. It introduces social media, Facebook, Facebook Friends and Facebook Time and their relationship with 23

44 political marketing. The chapter reviews prior research in political marketing, relating political marketing to social media in Malaysia and the Malaysian political scene. Concept definition of the measurement are covered along with deliberation on predictors of voters behaviour, impact of Internet utilisation on political marketing, Facebook and political activity, candidate s profiles, political attitude, party orientation, voter orientation and voters political activity. Finally, the chapter deliberates on the relationship and influence among predictors on voters behaviour, impact of Internet utilisation on voters behaviour, impact of candidates profiles on voters behaviour, impact of political aptitude on voters behaviour, impact of party orientation on voters behaviour and voters political activities and behaviour. The chapter ends with the conceptual development of the political model. One of the underlying theories in marketing is the Exchange Theory. This theory originated from the Social Exchange Theory, which explains the relationships between human beings in society. Exchange theory, on the other hand, explains the basis of relationships between sellers and buyers. Concepts from Exchange Theory in Marketing are then applied to the political context to become Exchange Theory in Politics (Figure 2.1). In this way, the application of marketing efforts and strategies is extended to the political arena. Although it may be considered a new field of study, political marketing as a research area has grown rapidly and has penetrated most aspects of political strategies and propaganda modes. The use of social media in politics has been an important development in this field in recent years. It has been successfully applied in the more developed nations. The question is whether social media such as Facebook can play an equally important role in garnering political support from the mainstream voters in Malaysia. 24

45 Prior to the discussion on political marketing, the scope of the study has mostly alluded to politics as the essential element of the research (Figure 2.2). The discussion on the application of politics to marketing covers previous research, particularly the application of marketing strategies in their electoral campaigns by political parties. Figure 2.2 takes a close look at the flow of the literature review process before having an understanding of the conceptual development and the establishment of hypothesis with regard to the voters decision model. Refer to (Butler & Collins, 1994; Glick, 1967; Kotler, 1975; Kotler, 1981; Kotler, 1981; Mauser, 1983; McGinniss, 1969; Newman & Sheth, 1985; Nimmo, 1970; Smith & Saunders, 1999). Figure 2.2 Review Flow and Conceptual Development of the Hypothesised Model 25

46 2.1.1 Definition of Politics By definition, politics is a progression in making collective decisions, encompassed with various social relations involving the authority or power subjected to particular procedures and guidelines (Wordnet Dictionary, 2013). Politics also applies to the behaviour of civil governments, corporate sectors, academic and religious institutions. Politics is fundamental in forming a workable organisation, in addition to establishing justice prior to ensuring peace within a country. Politics considers many aspects; the rights to choose, freedom of speech, possessing equality in practising one s rights as a voter, in electing leaders, as well as ensuring and improving morality standards in living. Another emphasis on politics is that it has always been about adopting manoeuvring methods and tactics when administering any given society, or running the state government, also envisioning that building and maintaining a community and establishing a platform to serve the community and its surroundings will be executed (Wordnet Dictionary, 2013) Definition of Marketing The complementary concept to be used here is marketing. Marketing is defined as a management process through which goods and services move from concept to the customers (Business Dictionary, 2010). Marketing emphasises the needs and satisfaction of consumers. Marketing is also defined as the activity of sets of institutions having the processes of creating communicating, delivering and exchanging offerings that have value to the consumers and other stakeholders (American Marketing Association (AMA), 2007). Marketing has been identified to consist the following activities: 1. Identifying what products/services are needed to fulfil demand 2. Determining the sellable price of goods / services 3. Strategising the distribution of products / services to reach the targeted consumers 4. Organising the right promotional activities. 26

47 The Chartered Institute of Marketing defines marketing as catering to the needs of the consumers and anticipating that the consumers can stay for a long-term (Marketing Teacher, 2000) Marketing raises public awareness. Therefore, sales can be increased, and if they do, the company will have a better reputation and a stronger presence in the consumer world. Furthermore, the philosophy of the marketing concept is that every firm should study its customers needs and try to satisfy those needs better than competitors do. Almost all companies have adopted the marketing concept one way or the other. Smith (1776) in his book The Wealth of Nations, emphasised that the needs of the producers should be considered in view of meeting consumers needs. This philosophy (although it was consistent with the marketing concept) was never adopted widely until nearly 200 years later. After World War II, many companies began to realise that hard selling was no longer a reliable way to generate new sales. Consumers had become more selective because they had increased discretionary income. Hence, they only bought those precise products that met their changing needs, which were no longer so obvious. So, marketers needed to examine: 1. What did customers really want? 2. Could the products be developed and provided to consumers while they still need them? 3. How consumers be kept satisfied? As a response to these questions, companies began to adopt a marketing philosophy, which involved the following: 1. Always start with the customers needs 2. Align the company s function with the need focus 3. Realise profit by capitalising on the customers needs over the long term. 27

48 To do the above most firms set up a marketing department with the main task of fulfilling the customers needs. These departments were often an extended version of the sales department, but today they are called marketing departments with a company-wide customer focus. Customers satisfaction has become the main objective of the department The Concept of Political Marketing As an outgrowth of marketing and the marketing concept, the political marketing concept was an inevitable development. AllBusiness (2010) defines political marketing as a tool devised to influence voters about certain matters, such as political issues, particular candidates for public office or public issues. Political marketing tends to adopt the same techniques as any other processes. The basic notion is to promote a concept or an idea, instead of highlighting a specific product or service, and to get people to support that idea. Political Marketing is actually the expansion of marketing concept as proposed by Kotler (1975) and his contemporaries. Henneberg and O Shaughnessy (2007) conceded that political marketing leaned closely on political candidates and was concerned with the manner in which voters political needs and support could be fulfilled. A comparison between the conventional products and services marketing with the marketing of political candidates revealed the difference in the promotional strategies used. Marketing of products and services focuses on the media to inform, remind and change the attitudes and behaviour of consumers. Political marketing should use similar strategies. Henneberge and O Shaughnessy (2007) also proposed that some other similar tools are used- these include market research and statistical analysis. It was later concluded by Butler et al. (2007) drawing a comparison between the marketing of goods and services and marketing of political candidates, that all marketing concepts such as 28

49 buyers and sellers, consumer behaviour, to name but a few, are equally applicable in political marketing analysis (Kotler, 1975). 2.2 Underpinning Theory Social Media The Internet World Stats (2012) indicated that the population of Malaysia stood at 29,179,952 of which 17,723,000 or 60.7% are active users of the Internet. It was also recorded that 13,389,250 or 75.5% of Internet users in Malaysia are Facebook subscribers. Kemp (2011) found that Malaysian Facebook users friending online activities were almost 80% higher than the global yardstick with one person having an average of 233 friends. It is also a striking fact that Malaysians would rather be surfing the Internet than watch television, indicating the fact that the time they spend on online activities has doubled. With that, it is plausible to see the country s Internet traffic congested by popular or trending social sites. Burson-Marsteller (2011), stated that if we were to refer to the monthly data, the Malaysian netizens visit and surf their social network sites 14 billion times every month (Table 2.1). For the age group of 15 and older (in total 11 million people) who have access to the Internet, 92.4% are regular Google sites visitors. Citing Nielsen (2011), Factbrowser also confirmed that the most active Internet users were from 20 to 24 years old, where almost 60% of them were regular Internet users and spent about 19.8 hours online every week. This easily contributes to the steadily increasing popularity of the social media in this country (Table 2.1) (Burson Marsteller, 2011). 29

50 Table 2.1 Top 5 Social Media Channels in 2011 in Malaysia (Source: Burson Marsteller, 2011) No. Social Media Population (000) % 1 Facebook 14, Blogspot 1, Wordpress Twitter Metacafe The latest phenomenon in communication is, arguably, the emergence of social media. The present study evaluates the role of social media as a communication tool to reach the audience, especially younger voters. The study by Ernestad and Henriksson (2010) concluded that the social media transition is bringing a rapid change in organisations mindsets. The most complex aspect is in the transition from traditional marketing strategies to social media marketing, as has happened in larger organisations. Wigmo and Wikström (2010) concluded that social media would improve brand awareness and reputation, and improve business-to-consumer relations. They also noted that social media could harness and control the nature of word-of-mouth between customers and potential customers. Nevertheless, similar to all efforts done in marketing, social media investments are not easy to measure as far as the Return on Investment (ROI) is concerned. Social media should today be used as a complement to traditional marketing and should be 30

51 used as a means to facilitate two-way communication. Poynter (2008) stressed that Facebook illustrates how brand and customer communication is changing Facebook The Concept of Facebook Communication has evolved as it moved from pictorials to speech to analog to digital media, but the strongest and most effective communications have always been a face-to-face conversation where both verbal and nonverbal acts play a role in developing relationships. The biggest barrier to relationships has always been distance. However, social networking sites (SNS) enable closer relationships without relying solely on proximity or verbal communications (Rau, Gao, & Ding, 2008). Facebook has been a leading SNS since its inception in 2004 by founder Mark Zuckerberg. It was the 4 th most visited website in 2010 (Bowe, 2010), with more than 800 million users as revealed in the studies by Elphinston and Noller (2011) and Marshall (2012). Facebook s success is due to its relationship-enhancing capabilities. A personal profile is required to register. The registration creates a user-driven page that can link with the pages of friends, exchange messages, post photos, create experiences, almost in real-time. Users can upload photos and use facial recognition to identify the individuals in the photos. Friends can be tagged for an emotional connection and illustrate affection (Moreno et al., 2011). In fact, one study found that over 57% of those posting pictures found themselves in romantic relationships (Mansson & Myers, 2011). In the context of this study, reviews of politicians Facebook pages revealed that Facebook functions as a social networking site used by politicians and political parties for political 31

52 purposes: to upload user-generated contents that include personal profiles containing their education background and political experience; photos of political activities; announcement on political events; replying to political issues; making election promises to woo voters; reminding voters on previous favours; and urging voters to come out and cast their votes Facebook Friends Social media has revolutionised the concept of friendship in social and psychological terms. For instance, with Facebook, to friend has become a verb in social media. In real life, both forging and ending friendships can take substantial effort, but on Facebook, it is as easy to make friends as to unfriend them. Friend has expansive connotations on Facebook, where a Pew Research Report by Hampton, Goulet, Rainie and Purcell (2011) revealed that 69% of the users use the word friend for high school classmates, extended family members, co-workers, college friends, immediate family, people from voluntary group and neighbours, while 31% of the friends in their Facebook (Facebook Friends) were almost strangers to them or barely knew them in person. A study on Kurdish lineage and tribal organisation by Costa (2016) revealed that Facebook Friends mainly consisted of acquaintances, with only a few of them being real friends. This shows that Facebook terms any type of social connection as a friend, lumping together longterm trusted relationships and the incentive-driven, superficial engagement based on, say, friending a retailer. Cronin (2009) noted the difference between traditional friend and a Facebook Friend. This definition of Facebook Friend indicates a value different from traditional definitions of 32

53 friends. The concept is also supported by Zywica and Danowski (2008). Facebook Friends in the context of this study refers to Facebook users who made online friending requests to politicians by utilising the add friend function made available through the Facebook platform. Politicians can accept or reject the friending request after reviewing the requestor s profile and political background. Once accepted as a Facebook Friend, both the politicians and Facebook Friends are free to interact and exchange information, news and opinion. Furthermore, the present study conceptualises Facebook Friends as one of the moderating variables on the relationship between predictors and voter behaviours Facebook Time Facebook keeps innovating new ways to engage its monthly active 1.5 billion users (as of 2015) (Wolfe, 2017). The average user spends at least 50 minutes on average per, or onesixteenth of their waking hours, on Facebook (Stewart, 2016). Ten years ago, the majority of active Facebook users were university students, according to research by Ellison, Steinfield, and Lampe (2007). Almost all (94%) of college students used Facebook for minutes daily and had friends. A study by Wiley and Sisson (2006) found that even in developing countries, 91% of students were engaged on Facebook on a daily basis. However, a Social Media Update Report (Greenwood, Perrin, & Duggan, 2016) published by Pew Research Centre arising from a more recent study of Internet & Technology Project, found that nearly 68% of adults in America were Facebook users. More than three quarters (76%) visited their Facebook page daily, with the majority of these (55%) visiting their Facebook page several times per day. This study also used the terminology of Facebook Time, as suggested by Lewis, Kaufman, Gonzales, Wimmer, and Christakis (2008) and Seib (2012), which determined the frequency 33

54 and types of interactions of connected Facebook users that could involve political marketing activity. Facebook Time in the context of this research refers to the amount of time politicians and Facebook Friends use the Facebook platform to read or exchange information, news, profiles, videos and photographs in order to generate interest, increase support or provide explanations on political issues. The amount of Facebook Time spent for political interaction can be measured in terms of hours spent on Facebook daily, weekly or monthly Discussion Figure 2.2, presents the conceptual model of the present study, which involves testing the moderating effect of Facebook Friends (Persons that politicians added as friends, upon receiving friending request in his or her Facebook account) and Facebook Time (time spent using Facebook during every Facebook viewing and interaction session). Hence, the existing belief of audiences can, most of the time, be reinforced by both brand and political advertising. This assures them that they can remain loyal since the values associated with the brand remain steady and true (Peng & Hackley, 2009). The level of user political activity depended on factors such as how much they are engaged with Facebook and their Facebook Friends political activities through the News Feed. These Facebook Friends are not necessarily their friends in real-life as they could have been someone they met once or someone who is a friend of their real-life friends. According to Vitak et al. (2010), factors that encouraged people to participate in political activities on Facebook depended on the length of time they had used Facebook, their political knowledge, interest, party affiliation, amount of exposure to political information in the Facebook News Feed, and general political participation. Vitak et al. (2010) also examined Facebook user skills related to politics. They believed that, apart from technical aspects of using social media, users also developed other aptitudes such 34

55 as civic skills. They stated that civic skills, particularly those developed in adulthood, were introduced as a complement to the more resources of time and money. Such skills were often thought to develop informal educational environments (such as a civics class) or in family environments. According to their study, civic skills also developed informally in adulthood through interaction with peers, and further examined Facebook as a potential environment through which to cultivate those skills. Regardless of whether occurring on Facebook or other sites, any political activity facilitated the development of civic skills, which in turn increased political participation. One advantage of the political activity enabled via Facebook was the opportunity to practice civic skills with a minimal amount of time and effort. In addition to the accessibility of Facebook at any time of the day, activities such as joining a political group or sharing a link can be accomplished with a few clicks of the mouse. These characteristics provided unique opportunities for participants to develop skills in their own time representing a lower threshold for informal civic-engagement education (Vitak et al., 2010). Online participation (for example content creation, wall posts, link sharing, commenting, and likes ) had been examined in previous research. Psychological empowerment associated with creating content on social media sites were related to activities, gratifications, and offline participation of users in civic and political processes (Leung, 2009). In the political arena, a report on Social networking sites and our lives (Pew Research Report, 2011) on a study of the 2010 US election conducted by the Pew Research Center, found that, when compared to Americans who were not active online, Facebook users who visited their Facebook account multiple times per day were two and half times more likely to attend a political rally or meeting. The Pew study also found that 57% of those active 35

56 Facebook users were more likely to promote their political choice, and 43% were more likely to vote. Comparatively, among the general Americans who are not active online, 10% of them reported that they had attended a political rally, 23% reported that they had tried to convince someone to vote for a specific candidate and 66% reported that they had or intended to vote. Therefore, the distinction and relevance of this study is using both Facebook Friends and Facebook Time as separate, but equally critical variables. 2.3 Political Marketing Niffenegger (1989) asserted that modern political marketing was not mere publicity and cunning advertising technique. It blended together each of the 4P s of the marketing mix, guided by marketing research with sophisticated segmentation and simulation techniques. Lees-Marshment (2001) concluded in her study of political marketing that sales-oriented and marketing-oriented strategies in marketing could be fittingly applied in politics and hence formed the basis of political marketing. Niffenegger (1989) regarded political marketing as the adaptation of techniques and concepts that were first pitched in business to facilitate the achievement of goals by political organisations. The exchange theory in political marketing (PMET) was first suggested by Kotler (1975) who outlined clearly that marketing was strictly concerned with how transactions came about, aided, encouraged and appraised as they serve as the exchange of value between the seller and the buyer. Kotler (1975) then conceived the marketing mix of the 4 P s in the context of politics. He equated product with promises and favours conveyed by political parties or candidates. To him, price was goods, and services was like electoral support for the voters, while customers were essentially the voters. Being the pioneer in Political Marketing, Kotler 36

57 (1975) put forth the idea that a promotional plan would encompass all possible tools or devices. These ideas are partly indicated in Figure 2.2. Figure 2.2 Review Flow and Conceptual Development of the Hypothesised Model 37

58 Henneberg and O Shaughnessy (2007) highlighted nine themes for further research with political marketing because they were distinct from commercial marketing practices that required careful modelling in concept and theories. Henneberg (2004) raised two specific questions. First, Is Political Marketing Theory based on the understanding of marketing activities normally referred to as Oxymoron? ; and second, Is Political Marketing Theory based on the Knowledge of Politics or Hendiadyoin?. The study also intended to explore political marketing existence and its epistemology. The findings showed that political marketing had established itself as an important subdiscipline of the mainstream marketing. However, stagnation in knowledge development was identified. Two different stances were identified in the core of political marketing theory as follows: a. A view that narrowed down the insight on marketing activities in politics b. A broader view of the holistic attempt to widen one s political knowledge. (Stephan Henneberg, 2009a, 2009b) Arndt (1985) identified four concepts that were regarded as pivotal in the ontology and epistemology of political marketing as follows: - A credible market environment - The exchange character embodied in political marketing - The structural link between political marketing and politics, indicating that ethical considerations were not dismissed (Henneberg, 2004) - The social embedding of a political system in other generic systems Furthermore, the methodology of the new concept of voluntary exchange can be used as the foundation and core idea to build a marketing theory since the underlying marketing theory 38

59 found that an institutional political paradigm in marketing could complement the prevailing microeconomic and marketing management concepts. The latter framework provides insufficient tools for analysing exchange structures and processes within and between organisations (Grönroos, 1997). A study by Lees-Marshment (2004) attempted to investigate the transformation of the government of the UK through political marketing revolution. The study used best party policy on key issues, audience figures for the Queen s Christmas broadcast and a political marketing literature overview. This was the first book to argue that marketing had permeated all areas of politics; not just political parties, but also local government, the media, parliament, health and education services, charities and even the monarchy. As such, it was an innovative thought, utilising a wide range of literature from two disciplines, political sciences and management, but it did not claim to be an in-depth tested empirical work. O Shaughnessy (2004), who conducted research on political marketing, stressed its media, complexity and turbulence and also spin, rhetoric and symbolism. Alongside both the control and interpretation of political marketing, he reported that marketing in politics could not simply be made similar to business marketing. He argued that political marketing could actually be harmful, as had happened in Canada and Britain. Hence caution must be taken when treating politics with marketing strategies. A psychological Exchange Theory developed by Foa et al. (1987) was used to examine: a) The structure of six different underlying resource categories; and b) The functional relations of these six resource categories. 39

60 The economic concept of resource constraint (scarcity) was also incorporated into the exchange process. The results were basically consistent with both the structural and functional relations derived from Foa s theory (1987). Moreover, resource scarcity was found to affect the types of resources an individual was likely to give in exchange. Miller and Lewis (1991) developed a value exchange model in their research to show the stakeholders approach to marketing management. The study provided tools to assist marketing managers in managing the marketing function strategically. The corporate value exchange model was shown to have provided the means to integrate and measure the direct value exchange relationship with the constituents, or indirectly among them, quantitatively. Miller and Lewis (1991) suggested the use of the measure in analysing the interpersonal and other one-on-one value exchange relationships. Social exchange in marketing by Bagozzi (1975) focused on three different scenarios: Reciprocity, Redistribution, and the Market exchange scenario. His findings showed that Exchange is a fundamental and universal aspect of human behaviour. Economic exchange models dealt with the buying and selling of material goods and services, while social exchange models broadened their scope to include social and psychological aspects of interactions. However, according to Bagozzi (1975), the social exchange model still exhibited serious shortcomings for the marketing scholar and practitioner. It was largely theoretical, unrealistic, narrow in applicability and lacking in its depiction of important facets of human behaviour. In light of these criticisms, the notion of an exchange system was proposed and illustrated as an explanatory framework. Finally, it was suggested that marketing could be viewed as a component of the social system, functioning as both a cause and consequence of social change. 40

61 Dant and Kaufmann (1992) found the dimensions of commercial exchange using the methodology of relational exchange theory, commercial exchange relationships, and transaction cost analysis. Their study found a method for measuring the structure of commercial exchange relationships, which was developed using contracting norms borrowed from Relational Exchange Theory. The norms were operationalised for a range of industry nonspecific governance structures and empirically evaluated. Results provided support for a multidimensional construction of the governance of commercial exchange relationships. A behavioural simulation was used in research on the structure of commitment in exchange by Gundlach, Achrol, and Mentzer (1995). The authors studied the effect of the credibility and proportionality of commitment inputs in an exchange upon the development of relational social norms, opportunism and long-term commitment intentions. They suggested that commitment makes the essence of successful long-term relationships, and they also investigated a series of longitudinal effects of the credibility of long-term commitment intentions, relational social norms, and opportunism in a one-time period on commitment inputs and long-term commitment intentions in later periods. Butler et al. (2007) traced political marketing theory building and focused on two specific topics: the political marketing research literature and the public sector marketing literature. The results showed that political marketing actually shared some common ground with public sector marketing even though this was not reflected in the current literature. They commented that the sub-disciplines were developed in mutual isolation causing constraints to their relevance and theoretical potential. For example, the political marketing research literature highlighted election campaigns and communications processes, while marketing for most politicians has involved public service delivery. On the other hand, public sector marketing was heavily influenced by the new public management that attempted to disassociate 41

62 administration and politics. The research highlighted the fact that the welfare effect of marketing activities in public finance and politics were profound, supporting the critical need for integration. In a study by Ruekert, Walker, and Roering (1985) on the implications of traditional organisation theory and transaction cost economics, they found that previous analysis of marketing within an organisation had focused on the coordination of marketing activities and on the organisational forms used to accomplish them, while ignoring the organisation of marketing tasks at the work unit level. They developed a contingency approach to the structure and performance of marketing activities at the work unit level, as well as higher levels within organisations by blending the theoretical implications of traditional organisation theory and transaction cost economics. Four propositions were developed to explain the effectiveness, efficiency and adaptiveness of various marketing organisational structures. In addition, four basic structures of marketing organisation were also explored. Bannon (2005) examined the internal marketing and political marketing uses of the methodology of internal marketing and stakeholder analysis and matrix of buy-in. His results were able to show that from the earliest political endeavours, politicians have acknowledged that the internal organisational support is vital as of the platform for internal and external electoral support, in spite of the fact that the internal marketing s appropriateness, limitations and efficiency has received a scant critical appraisal. Bannon first reviewed its appropriateness for political parties and projected the potential benefits and limitation. He also reviewed effective methods of the internal marketing process. Bannon also introduced the generic tools for assessing and improving the internal marketing process. 42

63 Calonius (2006) elaborated two specific topics to guide the study of political marketing. First, he discussed sociological research, and secondly, on Alderson s (1957) marketing research. Calonius s seminal work on the concept of promise in consumer behaviour and marketing had not been published earlier in any journal outside Scandinavia. Throughout the 1980s he developed a foundation for promises and management view of marketing, which were available in English only as conference presentations. It has only been during the past twenty years that his work has influenced the relationship marketing and service marketing literature. Now his work impacts marketing thought at large. Han (2007) projected on the marketing of politics by presenting a screening model in which the party maximises the political support of its members while individuals maximise their expected return. Additionally, he conducted a systematic political economic analysis of how political elite was influenced by, and influences, economic reforms in an autocratic country. He found that reducing autocrat practices could directly control economic resources and result in economic liberalisation that includes democratisation. This was done in post-reform China by altering the composition of the ruling Communist Party membership to keep the political and economic control. This resulted in the increase in educated individuals with greater private-sector opportunities. In his research on market segmentation and political marketing, Bannon (2004) had written extensively about the concept of segmentation, the development and application of tools in statistical analysis, disaggregation versus market segmentation and clustering (post hoc) segmentation analysis. He further discovered that market segmentation and targeting techniques were very much prevalent in the politics although the appropriateness, limitations and efficiency of market segmentation have not been given due attention and did not receive a proper assessment. 43

64 Menon (2008) stressed the implication of creating the model of voting behaviour in political marketing by combining the concepts of political science and consumer behaviour. He shortlisted eight significant functions of political marketing encompassing product, distribution, cost, communication, new management, fundraising, parallel campaign and internal cohesion management. Boettke, Coyne, and Leeson (2008) studied the assumption of government s failure to demonstrate how an economic approach to politics changed rather dramatically the way scholars approach the study of political decision-making. The persistent and consistent application of neoclassical models of economics also appeared to give an indication that once the full costs were accounted for, this failure was nothing but an illusion (Boettke et al., 2008). The authors also countered the arguments claimed to have been raised by Wittman (1997) who shed light on the underpinning economic theory and built an alternative model of political economy based on the Austrian ideas of the dynamic market process. Hughes (2007) introduced his approach to the methodology of branding theory to demonstrate that the brand offered a sustainable competitive advantage that can convince consumers to purchase a certain brand. His work showed that while there were statutory limits on Australian political advertising, the development of a political brand is allowed. Such a perspective would benefit political marketing practitioners to extend their marketing applications beyond a specific focus on advertising to one on branding. In the same vein, some theorists in branding had urged the development of specific theory and research in the various application areas of branding. Wiley (2008) argued with support from communication theory within the principles of relationship marketing. These principles included the CRM and e-crm, the narrow conception of marketing as a management 44

65 technology, the concept of externalities as spillover effects of economic transactions which brought social, environmental or economic effects and the challenge of new conditions of marketing. He presented a unique perspective on the relationship of value-sharing and buyingselling interaction. Satisfied customers get relevant, valued, fulfilled promises from their customers. Research by Hansen and Jensen (2007) focused on planned behaviour theory in a political context. The argument put forth is that being the service provider, the political party seeks to deliver some societal and other benefits claimed to enhance peoples lives, exchanging votes for the price. To offer support for this contention, data were gathered from a web-page survey and processed using the Structural Equation Modeling (SEM). The findings conceded that the theory of planned behaviour, with the inclusion of a path from perceived behavioural control to attitude, provide a good fit to the data to explain a high proportion (63.2%) of variation in the aspect of voting intention (Chiu, 2003). Yoon (1997) projected the endogeneity in political processes and further researched on the works of Ainslie (1974) on change preference. Buchanan (1993) attempted the interpretation of political process as an exchange of interests (and values) rather than the search for truth. Wittman (1973) projected how democratic policy was as efficient as the economic market and proposed an efficient hypothesis of the political market that recognised the fundamental difference between market and political processes. A market involves bilateral exchanges whereas the political process involves complex exchanges. The market for political services involves the question of legitimacy as well as efficiency. Wittman (1973) proposed an evolutionary theory of political system and efficiency that can be discussed within the given framework of constitutions that restrict political processes. Ricks 45

66 (1993) analysed the concept of the international political economic paradigm that played its role as a marketing channel tool. He stressed the importance of both economic and noneconomic political paradigms, which should steer the development of the channel marketing theory. Thus, a new theory of international political economic paradigms was proposed for the competitive international markets (Ricks, 1993). In his research on the generic functions of political marketing, Henneberg (2003) found that political parties used marketing instruments in their electoral campaign activities. It was examined and sorted out using the 4P marketing mix concept. These concepts have been subject to various criticisms. Henneberg (2003) adopted an alternative perspective called the functional analysis to describe the criteria for successful political marketing activities that catered for some political marketing instruments. Eight generic functions of political marketing were outlined as they proved relevant for varying exchange relationships and submarkets of the political sphere. A parsimonious set of interrelated dimensions was adopted in the analysis of political marketing instruments and activities. The emphasis was on the campaign-related political marketing management. Moufahim (2009) summed up that political marketing literature had primarily taken an instrumental approach; such as the election campaign, polls, and political communications among others. Moufahim concluded further that political marketing literature had borrowed many concepts from mainstream marketing. He gave a descriptive and normative recommendation for political actors on the best practices of marketing to be responsive to the electorate to come out victorious in elections. However, in scope, the theoretical development of the discipline was quite limited. The main focus in the deliberation was on the application of marketing mix-related concepts and tools by political actors and on comparative political studies. 46

67 Henneberg (2004) offered an alternative perspective that addressed the limitations provided in the study of Moufahim (2009) to apply the concepts of mainstream marketing in political marketing effectively. He generated a critical political marketing agenda, which survived on the existing strands of the political marketing literature. A recent study by Hughes and Dann (2006) about the direct benefit, value and management of voter s relationship concluded that the definition of commercial marketing had changed drastically. Political marketing also needed to be modified, adapted, and adjusted. They threw some light on the issues that had arisen from the new commercial marketing definition and the scope that had to be explored further. They stressed that further research was needed to address the interaction between political marketing and the stakeholder benefit, as traditional commercial stakeholder theory only applied to a certain degree of modification. As a whole, their paper established that political marketing had potential benefit from the new commercial marketing framework, with elements of political marketing currently being in closer alignment with commercial marketing than it was with the preceding definition. The principles of commercial marketing techniques were backed by the study results to aid a few matters- the political campaign process, the adaptation of stakeholder theory to political marketing, and requisition of components of quid-pro-quo value for value. Baines and Egan (2001) researched structure and strategy in political markets, which supported this idea. They used a few different approaches, such as the Oligopolistic Market Model, Monopolistic Competition Theory, and the Structure-Conduct-Performance Paradigm. In research by Blois (1974), he revealed that consumer-producer interactions had the attribute of mental stimulus processing services in political campaigns. In fact, candidates and political parties were aiming to deliver a special right of franchise bidding process to the government 47

68 by collaborating with the oligopolistic market. Blois further stressed the apparent flow of logic received from the market strategy, market structure and its performance. O Cass (1996) conducted research which looked into the political processes and electoral behaviour from a marketing perspective to lend some new insights into the political parties behaviour and performance. O Shaughnessy (1990) reminded that research and debate should emphasise on the parties marketing orientation. In a significant way, works on the actual marketing orientation of political parties were not deemed substantial yet. This research shed light on this issue by tackling the issue of marketing orientation of a major political party. Francois (2003) research stressed on the ballot technique in a voting analysis, where he studied the Austrians ranging from the entrepreneur to the political process. By differentiating the political process from the economic process, Francois highlighted the extent to which the political market process analysis was seen as both possible and relevant. The existence of information conveyance, which could be assimilated to a price system within the political process, with its subsequent profits, proved to be the motivating element for the political entrepreneur (Figure 2.3). Being doubly specialised and possessing all the Kirznerian functions of the Austrian entrepreneur, this entrepreneur also functioned as a coordinating agent in the political process. 48

69 Sales Strategy Political Marketing Strategies Marketing strategy Product strategy Facebook fans Social media Facebook Political marketing Votes Figure 2.3 Political Marketing Strategies and Votes Getting Research conducted by Fligstein (1993) on the market as politics, examined three ideas: the political-cultural approach; social market behaviour; and Japanese keiretsu theory. Fligstein linked the markets with politics to create a sociological view of the action in markets. He addressed the issue of how politics in markets work during the multiple stages of market development formation, stability, and transformation. During the market formation, actors in firms were seeking to create a status hierarchy that enforced non-competitive forms of competition. Political action is a resemblance to social movements. In stable markets, incumbent firms have stood by their positions against challengers and invaders alike. Within times of market transformation, invaders can reintroduce more fluid social-movement-like circumstances (Figure 2.3). 49

70 In a recent study, Pels, Möller, and Saren (2009) focused on the issue of marketing. They adopted the meta-theoretical analysis to make the identification of the contributions and limitations of the current research approaches to business marketing. Configurational approach for marketing offers a theoretical justification with respect to the empirical findings versus the relationship dominance dilemma (Figure 2.3). 2.4 Overview of the Malaysian Political Scene According to Ahmad and Ahmad (2009), the Malaysian 12th General Election (2008) witnessed a drastic change in the political scene, contrasting with the landslide victory of the ruling coalition Barisan Nasional (BN) in the 11th General Election. This was despite the fact that 84.5 % or two-thirds of reports and advertisements were related to the ruling coalition and only 15.5% covered issues concerning the opposition parties. Zakaria et al. (2009) reflected that, for example, in a by-election held at the Bukit Selambau Constituency, it was clear from the day of the naming of the candidates that the voting trend leaned more towards the Pakatan Rakyat (PR) coalition than towards BN (in terms of the voters socialisation, candidates and leaders, district, religion, ethnicity and age), although the latter had gone all out to win. The reasons for the change were complex. Ahmad (1999) concluded that the result of the election in Malaysia was a reflection of the complexity of the country s politics. Voting in the 1999 elections manifested ethnic affiliation in national politics. Lim and Wai-Mun (2008) concluded that the BN moral defeat could be likened to the eruption of a political volcano of accumulated dissatisfactions from various ethnic groups that the BN could no longer keep under control. Anwar Ibrahim, the former deputy prime minister and current de facto leader of 50

71 Parti Keadilan Rakyat, successfully united various opposition parties to form a multicultural coalition against the perceived communal coalition of the BN. Malaysian voters were orchestrated under the People Power Revolution against communal politics and the 3Cs of the cost of living, crime and corruption. Aziz (2008) observed that there was no concrete or strong statistical evidence to indicate or support the belief that exposure to the Internet led voters to vote for the opposition. Nevertheless, heavy users of the Internet are more likely to vote for the opposition parties. Those who have strong tendency to believe the contents of blogs and specific websites are more likely to favour the opposition. Voters located in the urbanised central zone (Klang Valley) are more likely to believe that Internet and blog sites disseminate accurate information, and because of this, they have the tendency to vote for the opposition. Balasubramaniam (2005) concluded that the euphoria surrounding the Reformasi movement and the struggle for greater democracy and social justice lost its momentum. Nasir and A. Selamat (2009) and M. H. Selamat (2009) determined that web mining had a low accuracy rate and it had to be improved for further experiments to get better results for finding relevant information on the web. In a recall graph, it was revealed that not every relevant web page was successfully retrieved and displayed to the user. The researchers used Artificial Immune System (AIS), web mining Internet technological tool to analyse the impact of social network in cyberspace about political issues in Malaysia to show how it could create an impact on the political situation there. Furthermore, Weiss (2009) pointed out that social media boosted the excitement of the polls and the opposition s odds, but they also posed new challenges for the future. Apart from 51

72 former Prime Minister Dr Mahathir Mohamad and several followers (who later left the ruling party, UMNO), the BN seemed unfazed by the results, even as Anwar Ibrahim schemed to lure enough East Malaysian Members of parliament (MPs) away from the BN to tip the scales in Parliament. Established political parties have limited space to manoeuvre ideologically, lest they alienate their core constituencies, while activists operating within civil society are less constrained. Politics Political Marketing Political Marketing strategies Face book Face book fans Figure 2.4a Political Marketing Strategies and Facebook Fans Samad (2001) observed that the political and socio-economic scene in Malaysia had been affected by the emergence of IT and Internet. In early 2000 Malaysia, under Prime Minister, Dr Mahathir Mohamad appeared to be standing on a double-edged sword. The paradox was while Dr Mahathir's administration was a strong advocate of Information Technology (IT) and Internet in the interest of economic development fearing the erosive effects on political control and influence. IT and Internet were both a dream as well as a nightmare. ICT could have a profound impact on political development, in the opening up of new avenues for expression of popular dissent (Figure 2.4a). 52

73 Butler and Collins (1999) concluded that political marketing is a taboo in Malaysian society. Another observation was that the type of services normally connected to the political influence and constituency activity showed a convergence of politics and public sector service provision. The formal campaign in the political marketing debate had definitely been overemphasised at the expense of a more inclusive perspective. 2.5 Measurements in the Research There are striking similarities between voters and consumers in affective, cognitive and behavioural responses to advertising as political parties, and commercial entities apply similar marketing techniques in certain circumstances (Peng & Hackley, 2009). Although there is the possibility of overlap between the voters and consumer s role, there is an apparent contextual difference between election and consumption as asserted by Lock and Harris (1996) with attendant differences in attitude formation and decision-making. P. Kotler (1999) and N. Kotler (1999) asserted that the application of brand marketing management conceptual frameworks could deepen one s appreciation of voter attitudes and behaviour in the political context. Peng and Hackley (2009) concluded that there is a dearth of empirical research that probed the voting-consuming analogy shedding light on a specific segment of voters and their responses to particular political marketing stimuli. Marketing polling was used as the basis for the Political Marketing in the voter behaviour model, in the study done by Ben-Ur and Newman (2010), with a model assumption that the people who vote during a campaign are considered consumers of the service offered by the politicians. In fact, candidates are chosen by voters based on the perceived value that the politicians will provide for them if elected. 53

74 Five different cognitive domains that drive voters` behaviour are presented in the voter behaviour model. The main proposal behind the model explains that one or a combination of the domains will be relevant in a given campaign. The combination and importance of each domain will be different in each election. The model includes the following five domains: i. Political issues where candidates advocate and promise to enact if elected to office. The issues gain voters` attention, and would normally revolve around issues that people feel in their pocketbooks (El-Khoreiby, 2012). ii. Social imagery the stereotyping of the candidate tends to exude appeal to voters making it relevant to some particular segments of the society (El-Khoreiby, 2012). The candidate or political party build an image in the voters minds on the basis of their associations with the image being either positive or negative in the voter`s mind, depending on the group the candidate is trying to associate with. iii. Candidate s personality - The use of words and pictures is reflected in the personality of the candidate but in a different style. Here, relying on personality traits, the candidate (or political party) will attempt to reinforce and build an image in the voter s mind iv. Situational contingency - This dimension represents how a voter can be influenced by hypothetical events presented by competing parties or candidates during a campaign. At times, getting a small percentage of voters to switch their loyalty to another candidate or party means winning the election. v. Epistemic value - This dimension deals with the extent to which a voter is curious or novel in choosing a candidate. 54

75 2.6 Predictors of Voters Behaviour The Internet and Political Marketing The Internet has made creating news easier than ever before (Norris, 2004). Arthur (2010) stressed that in the 2010 election in England, the Internet was a more prominent tool than news media. According to 2010 surveys, 60% of British adults surfed the Internet every day (Office for National Statistics, 2010) and almost 50% of the British used online sources to gather information in the course of the election campaign. The UK 2010 General Election was regarded as the Internet election because Internet carried an increasingly significant role compared to previous elections. The outcome of the election was determined by events in the broadcast media, or more specifically, the three leaders debates (Ormrod & Savigny, 2012). Johnson (2002) believed in online and web-based surveys because they are a fast and inexpensive way to reach voters, with the promise to become increasingly popular. There are three advantages of conducting an online survey: first, there is no cost of delivery and printing, and the population can be very large; second, since the responses are keyed in electronically, it is faster, and data errors and costs are reduced; and lastly, data is instantly ready for analysis (Kotler, 2006; Ray & Tabor, 2003). On the other hand, in a web-based survey, samples are self-selected, and the results cannot easily be generalised to the larger population (Birnbaum, 2004; Schiffman & Kanuk, 2004). There are other criticisms about online surveys related to survey research methods in general, some of which are having ill-defined objectives, unclear targets, inappropriate sample sizes and representation, biased questions and lack of distinction between correlation and causation as found by researchers Smith (2003) and Crawford (2004). The greater likelihood of defective surveys online may arise as the result of relatively new and fast-growing 55

76 opportunities for online surveys that sometimes encourage less experienced researchers to adopt the method. Furthermore, after Miller (2001) examined sampling and non-sampling errors related to online research, particularly sample representativeness and measurement validity, he expressed concerns related to computer hardware and Internet service providers lack of standardisation, which may affect the way the respondent sees and reacts to the survey. Although this deficiency can be adjusted with propensity scoring, Miller stated that online and traditional survey respondents often vary in reporting key demographic variables. According to him, the sources of non-sampling errors were: lack of measurement, control and standardisation; limited access to online media; respondent refusals; non-responses to questions; inaccurate responses; and incomplete survey responses as supported by the study of Ben-Ur and Newman 2010). Miller suggested two ways to increase sample representativeness: namely, using mixed-mode surveys, and gathering data across a variety of media. A comparison of the characteristics of online users in 2001 to the general population is represented in the 2000 US population census (Smith, 2003). He found the online population to be wealthier, younger and has a higher percentage of males and married individuals than the 2001 online population. However, the gap between the demographic characteristics of the two populations seemed to be diminishing, as Internet access became more widely available (Ben-Ur & Newman, 2010). Certain respondents fit the research needs better than others since the online population is a more specific population. Arnold (2004) stated that fitness was usually concerned with the subject area of the research such as computer ownership or research on technology usage. 56

77 However, since there was no clear evidence of the difference between online respondents and the rest of the population in terms of the level of political involvement, this difference did not seem relevant to research in political marketing. Ben-Ur and Newman (2010) concluded that there was strong evidence that there were very limited differences in voters` perception and only a slight difference in voting behaviour, although some differences were shown to exist between online and mail respondents demographics. Taking into account the claimed limitations, it was suggested that online surveys might be a practical and productive method of obtaining data. Furthermore, a more accurate prediction of the election result, which was close to that of the real result was provided by both online and hard copy questionnaire surveys, especially the paper surveys, compared to the performance of standard political polling methods Facebook and Political Activity Esposito (2012) stated that there were no boundaries for politics and political issues on Facebook. Their impact is known for their ability to create online public spheres for discourse and citizen engagement (Hand & Ching, 2011). Political campaigns created official fan pages across the U.S. for their candidates or auxiliary groups in These pages informed the public of the campaign s activities, about who the candidate was, what they were doing, and about their stance on policy issues (Esposito, 2012). According to Esposito (2012), the first campaign that made effective use of the Internet was the 2008 Obama campaign. It was not an easy task, as it had to cover vast volumes of communications in an environment rapidly evolving, both in terms of technology and the life and death of sites. Due to the very nature of the web, it cannot easily be managed by large corporations, let alone by political parties. Moreover, success would mean the outcome of accepting diversity (or even anarchy) rather than controlling it. During the campaign, there 57

78 was a heavy decline in membership of political parties. At the same time, there had been a decline in the social networks that link politicians to wider communities. It had not yet been established whether or not social networking sites offer much potential to politicians and political parties. Meanwhile, political blogging appeared to have a substantial effect. In the UK, the leading political blogs seemed to be conservative leaning. According to the number of hits that the leading blogs were getting, it was obvious that they were reaching considerably more people than traditional political magazines. Some new concepts and theories had to be developed to deal with virtual communities and communication, specifically those used for activist advocacy and communications. Burns (2009) pointed out that on the 60th anniversary of the People s Republic of China in October 2009, the nations QQ social networking site overtook the Facebook platform by reaching 330 million users. As Harris & Lock (2010) noted, if it were a country, QQ would be considered the fourth most densely populated nation on earth Candidate Profiles There was an emphasis laid by Smith (2009) that these days, candidates have the ability to promote themselves, talk and explain their policies and agenda and interact with voters in different ways than they did in previous elections. Westling (2007) argued that compared with campaign website ability, Facebook is superior in integrating political communication into a broader social network. If a candidate s website can be analogous to a campaign rally for which a college student can be accountable, a candidate s Facebook page can function in much the same way as a campus-based campaign rally. Fagerlund-Savisaari (2010) observed that politicians used Facebook mainly for networking and campaigning. They either did not know, or their parties did not have common guidelines for Facebook usage. Politicians updated their profiles themselves as Facebook was considered a useful tool. The reasons for 58

79 joining Facebook varied from curiosity, interest in social media, networking purposes and being invited by a friend. Some politicians joined Facebook only for campaigning purposes because it was considered as essential for a politician to have a profile (Miller, 2006). Miller (2006) added that websites could be manipulated to access voters who would not normally watch television for a debate, for instance. It was a challenge for candidates to control the kinds of information spread on the web. Both supporters and opponents would be able to create profile pages for candidates without having to ask for their permission. It was also very likely that random people could post messages that might be offensive to others (voters or those supporting the opposition parties) on the pages. To help predict and explain voter behaviour Ben-Ur and Newman (2010) conducted a study the purpose of which was not only to measure voters intention and predict the votes but also to allow the political marketing strategist to understand voter motives and preferences so that the advantages and disadvantages of their candidate was known to them in the political marketing at the time. With this knowledge, they could take action to reposition the candidate to generate a better competitive position and a better voter response. Demonstrating competitive advantage was one of the major components of the strategy (Cosgrove, 2012). This included positioning to help distinguish the political candidate from current and potential competitors. Positioning was successful when it built on the competitor`s distinctive competence or strengths (Walker & Mulline, 2008). The market and its buyers, i.e. voters, whose values and aspirations drove their perceived needs, were the focus in such positioning. The success of the marketed entity was assured by answering these needs better than other competitors could 59

80 The dimensions in which consumers perceive competitive offerings can be classified into three categories: simple physically based attributes, complex physically based attributes and abstract attributes. The perceptual attributes are subjective and are considered the basis for the positioning strategy, although they are influenced by physical characteristics (Walker, Boyd, & Larre, 1999). For example, the notion of electability is the result of the candidate`s physical and behavioural characteristics in the U.S. political reality. As Paul Tsongas was not physically appealing, the Democratic presidential candidate was considered not electable by women voters in the 2000 U.S. presidential elections. Ben-Ur and Newman (2010) believed that once the elements of a market, competitors and evaluative dimensions were identified, the data collection and analysis would generate the perceived image of the marketed entities; in their case, the competing candidates for the presidency in the USA. Newman s (1981, 1999) model, adopted by Ben-Ur and Newman (2010), explained the voters cognitive domain with five major dimensions (constructs). Ben-Ur and Newman (2010) followed a process of data collection and analysis to create an image for each candidate in the U.S. presidential elections that could be studied and manipulated to better their competitive position. Johnson (1971) stated that when the image was closer to the ideal for a sizable segment of the market, the repositioning strategy increased market share, differentiated it from other products it must compete with, and in dimensions weighted heavily on consumers preferences. This scheme encompassed actions taken to minimise the effect of a negative image or strengthening positive characteristics that were weak in the promoted candidate. In order to improve a position for a product, Boyd, Ray, and Strong (1972) proposed that a marketer could: 60

81 Affect those forces which had a strong impact on the choice criteria used for evaluating brands belonging to the product class Add characteristics to those considered salient for the product class characteristics Increase/decrease the rating for the salient product class Change the perceptions of the company s brand with regard to some particular salient product characteristic Change the perception of the competitive brand with regard to some particular salient product characteristic. Some of these strategies were used by Ben-Ur and Newman (2010) to suggest repositioning of the candidate s image as well as to manipulate the opponent s image to improve the strategic position of each candidate. Bluhdorn and Szarka (2004) concentrated on small parties and especially the Green parties of France and Germany. It was then concluded that when changes take a more rapid pace within the political systems, the consequences of candidates ability (or the lack of it) to recognise and systemise their strategic marketing positioning choices were not seen as significant. Furthermore, small parties strategic (marketing) capacities were analysed by Bolleyer (2007) as being to realise goals and to identify the relative importance of their party system and corresponding positioning. The last category of critical success factors related to the e- marketing competencies and was considered to be the most contemporary one. The increasing importance of these factors was because of their multi-channel public servicing potential, their facilitation of improved service, cost minimisation, deeper relationships, better communications and customisation. The primary function of e-marketing was to complete and expand the choice of the channel rather than replacing the existing range. 61

82 According to Thrassou, Vrontis, and McDonald (2009), e-marketing ultimately benefitted from offering added value both through improvements to actual and perceived offered value. E-marketing which focused on family brands or corporate identity clearly showed some commonality with brand advertisements. Johnson-Cartee and Copeland (1991) in a study related to advertising, regarded hope and glory as a form of negative advertising when it came to looking into the opponent s suitability for office with regard to their character or previous performance. Negative advertising, in both political and commercial arenas, remained rare in Britain since that time (Peng & Hackley, 2009) Political Aptitude Vitak et al. (2010) examined Facebook users skills related to politics. They believed that apart from technical aspects of using social media, users would also develop other aptitudes like civil skills. Regardless of where it appeared on the World Wide Web, any political activity would be able to help cultivate one s civic mind and this automatically increased political participation. The political activity via Facebook exposes us to a striking advantage that it is the opportunity to practise civic-mindedness with the least amount of effort and time. Other than the accessibility to Facebook, activities, like being a member of a political group or sharing a link, can be carried out with a few mouse clicks. Vitak et al. (2010) stated that these characteristics provide great opportunities for voters to develop their own informal civic engagement skills in their own time. Haron and Mokhtar (2010) argued that a lot of candidates and political parties lose sight of the market due to the lack of systematic discipline, resulting in poor decisions and various political marketing mistakes. 62

83 2.6.5 Party Orientation Solomon, Bamossy, and Askegaard (1999) and Peter and Olson (2005) offered a new insight with regard to the cognitive processes and behaviour in commercial consumers and media audiences in their management literature. They asserted that marketers use different combinations of the 4Ps to try to influence these variables. Vakratsas and Ambler (1999) concluded that if consumers do not find media messages personally relevant, they are likely to disregard them, whether political or commercial. Lees-Marshement (2001) agreed with the idea that political marketing applies core marketing concepts and techniques which are then adapted to ensure that they are compatible with the traditional tenets of political science. Moreover, she supported the idea that the most successful political interest groups are now resorting to political marketing to create the package they offer to win the hearts of their supporters, using a four-stage process of demonstrating market intelligence, creating their product, relaying information to supporters and stepping up the campaign progress. Lees-Marshement (2001) also created a marketoriented party model to describe how political parties utilise the whole range of marketing techniques and concepts by incorporating a range of behaviour. Stromback (2007) established that the market-oriented approach was applied within the scope of the media system and the political system. He further pointed out that the reason some parties choose to be market-oriented or sales-oriented can be justified by weighing on the differences between countries or between parties within countries. Dann, Harris, Mort, Fry, and Binney (2007) who examined the position of political marketing in the wider marketing field, strongly believed that political marketing allows relevant parties to rejoin commercial marketing and return the theory and practice back to the major 63

84 discipline. In other words, the development of unique frameworks is a key to the future of the marketing discipline (Thrassou et al., 2009). According to Curran and Seaton (1991), as newspapers became more widely available throughout the nineteenth century, political marketing and market segmentation and target segmentation were ways to define what customers would want or accept from a product, service or idea, at what price (cost) as well as the best way to reach them Voter Orientation The basis of the political marketing concept is target audience/voter identification. Parties are required to understand more about their main and potential voters if they have serious intention to fulfil the needs of voters. It is not a strange thing to include images of the target audience in commercials, where it is widely accepted that advertising depends on consumers self-identification with advertising images to make their involvement with a message firmer (Robinson, 2007). Lafferty and Hult (2001) defined voter orientation (VO) as an orientation towards the customer which was integral to market orientation research. This framework can well be applied in other countries or their electoral systems, although at the time of writing, it had never been applied in the mixed member proportional electoral system such as New Zealand. This research suggested a possibly useful way of linking political advertising content with the electoral outcome. The framework did not apply to the political advertising or political marketing fields; if it did, it could perhaps strengthen the good impact of the utilisation of political marketing in relation to voter-and-media-generated election results and may well provide a very useful tool when it came to the analysis of the effective power of advertising and campaigning in future campaigns (Harris and Lock, 2010). 64

85 2.6.7 Voters Political Activity Westling (2007) believed that candidates would probably continue actively using Facebook if they know they would be running for office again in the future. Similarly, for the general population, Facebook users` political participation would likely be increased. Peng and Hackley (2009) dwelt closely in the market and consumer research, strategic market planning, positioning, segmentation and targeting deployed by political marketing campaigns, although it was a fact that these techniques are greatly invisible. Practitioners in political marketing mostly used advertising and other marketing communication techniques in various aspects. Khatib (2012) observed that one of the most important tools in modern elections was the opinion poll. He concluded that to run a successful political campaign, one of the prerequisites is the incorporation of opposition research into the candidate s campaign to find out about their opponents. 2.7 Conceptualised Relationship and Influence between Predictors on Voters Behaviour In normal circumstances, candidates would attempt to connect with voters using the more classic means of communication such as the television, radio, newspaper, magazine, posters and seminars. The Internet has been very useful in providing means of communication in a political promotion. Wattal, Schuff, Mandviwalla and Williams (2008) concluded from their observation that one of the main factors why Obama won the U.S. presidential campaign in 2008 was the use of the Internet The Impact of Internet Utilisation on Voter behavior Chu and Kim (2011) asserted that the expansion of the social media aided the consumers as they could explicitly communicate with each other via various advertising messages. J. Park and Choi (2011) and S.M. Park (2011) reiterated the important role of social media as a new 65

86 paradigm for communication, which led to some surprising changes in the political and social landscape all over the world. Although a candidate having a very large number of fans suggests that he or she is very popular, thus standing a higher chance of getting elected, the quantity does not always guarantee success. Esposito (2012) introduced as an example Barack Obama s campaign page as an example, which became the largest Facebook page on the network during the 2008 election. Esposito noted that by the 2010 midterm election online campaign presences were firmly established in the political process. However, he pointed out that during the 2010 gubernatorial election in Texas, USA, the two central candidates gathered, in total, more than 203,000 fans on their two pages: Bill White had more than 153,000 fans, and Rick Perry had more than 50,000 fans (Facebook, 2010a, 2010b). However, Rick Perry was re-elected Governor of Texas with 54.97% of the vote, while Bill White garnered only 42.29% of the vote in spite of the fact that White attracted triple the number of Facebook fans as Perry did. The outstanding popularity and rapid growth of Facebook encouraged politicians to rush into using it for political promotion. Furthermore, Facebook has shown that it is possible for individuals to create groups to share some common interests or to form competence groups or even groups with the same beliefs and views (Yousif, 2012) The Impact of Candidate Profiles on Voter Behaviour Some candidates included long biographies and resumes in their Facebook pages such as Senator Jim Web in 2007; conversely, others like Webb s opponent, Senator George Allen gave only basic contact information in his Facebook profile in Gulati and Williams 66

87 (2007) reported that candidates who posted in their personal profile could attract as many as 2,429 supporters, while those who did not, were unable to attract more than 429 supporters. The politicians appearances were especially emphasised by voters; they were also particular when it came to commercial messages that featured celebrities. According to Kotler, Armstrong, Saunder, and Wong (2005), advertising practice found some similarities between an advertisement that projected politicians personal profile and the endorsement type, both of which were used to deliver a brand message in the commercial context. If such a link could be established between the two domains of trust, namely the interpersonal and the incumbent, then there was also a chance to identify citizens-voters who were open to a common perspective with candidates. It had been observed that there is a strong relationship between interpersonal trust and incumbent trust. This suggests the importance of public officials trying hard to identify the issues critical to the citizen-voters they serve and to clearly identify the most important shared issues of agreement since these issues are likely to strengthen the concept of mutual trust. On the other hand, because of the relatively weak relationship between interpersonal trust and political cynicism, many candidates believe that that trust as a factor is not sufficient to convince the more cynical voters. Regardless of how sincere a candidate is, the cynical voter is most likely to dismiss or downplay such declarations of trustworthiness due to their cynical nature. The result of this research is useful to academic researchers, campaign managers, and political strategists who are seeking to discover the importance of trust factor in political strategy, tactics and campaign development in order to influence voters behaviour. 67

88 In an article written by Davies and Mian (2010) The reputation of the party leader and of the party being led, they studied the similarities between the popularity of political leaders and that of political parties and concluded that causal links could be detected between them. This, in turn, highlighted the role of the leader s reputation in a political party. The authors further argued that a change of leader would definitely result in a change in the party s reputation, the success of the party and the electoral outcome. In comparing their 2001 and 2005 data, Daviews and Mian (2010) found the unexpected influences that a change in a leader s reputation appeared to be contingent on the voter s willingness to vote for or against the party being led. The research findings emphasised that changing a leader was more critical to the attractiveness of the party concerned and to motivate supporters to become members than attracting new voters or benefiting its opponents. Harris and Lock (2010) suggest that the study made an important contribution to establishing a political reputation as a formal construct. Esposito (2012) explained four needs proposed by Katz, Gurevitch, and Haas (1973) to describe why people would use Facebook political candidates fan pages. The needs are cognitive needs, affective needs, integrative social needs and personal integrative needs. A review of candidates pages showed that these sites were rich in information about the candidate, political events, activities, issues and other campaign-related information. Because information was released continuously and aggregated on the page s timeline, users, motivated by cognitive needs had no shortage of content. Affective needs are those related to the satisfaction of feelings or emotions. Users who seek a sense of belonging or have emotional attachments to political candidates may feel satisfied with connecting to and participating on these pages. 68

89 Users may, for instance, participate in online electoral activities because they feel an emotional connection to the candidate that motivates their online participation. Personal integrative needs are related to improving personal status or credibility. Candidates taking opportunities to create content by sharing their thoughts and opinions on particular issues would resonate with a user who is motivated by personal integrative needs. When this need is satisfied for the users, they may participate more frequently as a result of enjoying the status they receive from being participants in online discussions, or in order to serve in a role as a subject matter expert within their own social network. Social integrative needs have the tendency to be satisfied through connecting with friends, family or other likeminded individuals. The function of social websites like Facebook is to meet the needs of these kinds of users. They may participate because they learn what their friends and family are doing or they may be looking to meet new people. From the study of Esposito (2012), it was found that determining the kinds of needs that are satisfied by using a candidate s page, as well as to predict how those needs will affect the type of online participation behavior is quite possible The Impact of Voters Political Aptitude on Their Behaviour Ever since the emergence of mass communication technology, the public is no longer confined to biased sources of information. With voters education, they became more critical in their decision making process and less dependent on parties. (Alexey, 2006) made the assertion that the pattern in electing candidates has become similar to purchasing goods or using services, something in which they are already very good at. Vitak et al. (2010) concluded that social networking sites such as Facebook had enabled users to share their political beliefs, support their respective candidates, and interact with others on 69

90 various political issues. They found that the strongest motivations for people to use online political information sites were to seek guidance about who to vote for and what the issues were and information gathering for more purpose than mere entertainment. They also argued that the Internet has always been a practical source of political information. The use of cognitive or information gathering needs extended the uses and gratifications theory, a theory used to determine why people choose to use specific media. (Ancu & Cozma, 2009; Kaye & Johnson, 2002; Leung, 2009; Park, Kee, & Valenzuela, 2009). Many studies have explained the theory further by applying it to blogs, i.e. websites where individuals publish their opinions, ideas, or information on a regular basis. Studies have also sought a nexus between consuming the websites content and voting behaviour, in addition to clarifying the uses and gratifications needs for reading political blogs (De Zuniga, Puig-I- Abril, & Rojas, 2009; Eveland & Dylko, 2007). Lewis (2011) claimed users who read blogs associated with online political discussions are more likely to vote than non-readers. Online participation in terms of content creation like wall posts, link sharing, commenting, and likes were also examined in previous research. According to the literature, psychological empowerment associated with creating content on social media sites is related to both the uses and gratifications and the offline participation of users in civic and political processes (Leung, 2009). Leung (2009) explained that online content creation was motivated by needs for recognition, cognitive enrichment, social interaction and entertainment and a correlation existed between all the uses and gratifications and content creation. Leung s results showed that the more content a user created, the more that recognition and social needs were gratified. Leung measured user-generated content in time per day spent on content creating activities like personal web pages, blogs, forums, posting videos on YouTube and contributing information on Wikipedia. Similar assumptions could be drawn about the satisfaction of 70

91 personal integrative and social integrative needs through creating content on Facebook, although the Facebook content was not measured specifically (Esposito, 2012) The Impact of Party Orientation on Voter Behaviour The implication of the contextual dimension for political marketing communication is substantial. Marketing communications focus has shown great changes through the new symbiotic relationship with the former being frequently predominant, especially for offerings such as political ones, whose added value is largely subjective and abstract. This, according to Servaes and Carpentier (2006) has created a major effect on the practice of political marketing communications: shifting the focus from function (i.e. awareness, knowledge, information, etc.) to perception. Overall, based on the findings, there is a serious demand for political marketing communications on intervention by which voters perceive their environment, towards dynamic shaping perceptions and expectations. Thrassou et al. (2009) asserted that the latter two factors naturally lead to increased public satisfaction. Khatib (2012) concluded that due to Jordanian parties failure to apply the principles of new marketing concepts in running their political campaigns, there was no significant statistical effect of adopting a market orientation strategy. This result, however, is in contradiction to the findings of some researchers (Lees-Marshment, 2001a, 2001b, 2008; Newman, 1999; Ormrod, 2005). The finding was very important for political parties, and it was necessary to explore more thoroughly and explain why Jordanians did not believe that their parties made enough effort to understand their wants before designing their political campaigns. They did not seem to believe that parties objectives were to discuss the voters main issues and that parties programs were developed based on voters opinions and wants. 71

92 Moreover, they did not believe that parties programs were designed to meet their demands, nor that those programs could satisfy their needs and wants. From the results of the study of Khatib (2012), it appeared that in Jordan, the voters believed that politicians do not have a market orientation. Instead of developing programs and products targeting voters wants and needs, politicians proposed their own products and tried to convince voters that they were right. Voters interact with the representation of the party through the campaign message, but finding a way to relate the message to electoral outcomes is challenging. Robinson (2010) concluded that the market orientation might be related to the achievement of electoral objectives by studying the messages embedded in party advertising. Parties may even include personal pronouns like I or We in their advertising campaigns to let voters feel united with them. The need to satisfy relationships with existing voters is a fundamental political marketing concept. The reason behind this is to maintain a core group of supporters who may be relied upon to vote for a party. Robinson (2010) suggested that, if parties want to maintain a consistent relationship with their core voters, they should project consistent leadership messages in their advertisement and evoke their history and myth. They should also provide evidence of meeting previous promises. This is actually a trade where voters exchange their vote for the party or candidate that is capable of offering the most value and can best meet their needs. In political advertisements, the offer can be perceived from billboards and newspaper advertising, briefly stated in a few words, a slogan, logo or photographic image. Robinson (2010) looked for offers that depict something new or different related to the competition. This may have been words or images signifying added value to existing policy platforms; particular benefits that the party would 72

93 provide in exchange for the party vote; or something new addressing an unmet need in the marketplace. Political Engagement as far as campaign managers standpoint on political engagement is concerned, is associated with two key components, voter mobilisation and campaign participation (Campbell & Kwak, 2010; Dale & Strauss, 2009; Gillespie, 2010; Michelson, Bedolla, & McConnel, 2009). The former is often measured either by actual voter turnout or through the intent to vote. The latter refers to the level of engagement a potential voter has with the campaign either through volunteer work or participation in their events. According to research findings, campaign activities like door-to-door surveys (Gillespie, 2010) or phone banking (Michelson et al., 2009) can impact voter turnout or intention to vote. Michelson et al. (2009) believed that phone banking, which is a popular campaign technique that relies on technology, has been shown to have an impact on voter response (Esposito, 2012). Gueorguieva (2008) and Robertson et al. (2010) contended that two key components in voter mobilisation are primacy and recency. The former refers to the salience of the issue or candidate to a potential voter, and the latter refers to the time component surrounding the act of voting, suggesting that time has an impact on the effectiveness of a message or call to action Voters Political Activities towards Voter Behaviour The level of political activity of Internet users depends on how much they are engaged with Facebook and their political friends activities. Vitak et al. (2010) outlined some factors that may encourage people to use Facebook to communicate and join in the political activities such as political knowledge, efficacy and interest of political parties. 73

94 Alsamydai and Rudaina (2011) stated that the feasibility of success of any candidate or political party would rely greatly on how well they use an effective and influential promotional strategy. Automatically, the political promotion activities are now an essential and an effective component used by candidates and political parties to influence voters and their attitudes, convictions and views. Promotional activities build a positive mental image of the candidate and subsequently, get voters to vote for them. They also claimed that political promotion factors such as advertising, personal contact, and public relations are important in delivering information to the target audience to influence individuals in making a choice of a candidate. The effort to create a positive image of the candidate thus depends on these various marketing efforts to urge and convince voters to vote for the candidate as the information is well disseminated (Jo Ander and Michael, 2011). Jo Andera and Michael (2011) pointed out that political candidates must attempt to convey positive image through their activities and campaigns rather than creating a negative image. Owing to the fact that voters have heterogeneous preferences, milder positioning differences are vital, and thus, election campaigning possesses an innate marketing character with multifarious similarities, rather than differences, in salesmanship in business and political domains. Commercial and political advertising overlap each other in terms of function as well. Hackley (1998) argued that the function of advertising is often to reinforce rather than to persuade. Political advertising is very similar to brand advertising. It is seldom possible to persuade individuals who are deeply brand-loyal to change. The main target achievement is to make audiences remember the brand, so it has to stay current, topical and active, at the same time keeping intact a sense of visibility and presence in the marketplace. Ansolabehere and Iyengar (1995) contended that this is also true in political as well as commercial advertising. 74

95 Yousif and Alsamydai (2012) claimed that in their study, the respondents positively received and dealt with promotional messages posted via Facebook. It seems though candidates and political parties follow every step carefully in order to gain voters` support, sometimes they fail to do so. Mistakes are inevitable in any process, and political activities are no exception. Haron and Mokhtar (2010) have identified the following mistakes committed by candidates and/or parties: They neglect inconsistency of voters tastes and continuously changing social and economic circumstances. They also did not have a systematic and continuous market research platform. Their ignorance of voter perception that leads to an insufficient understanding of voters behaviour, thus leading to inappropriate decisions. Political parties tended not to pay attention to the essence of voter behaviour, resulting in unreliable predictions and leading to an erroneous approach. They carried out needless practices without significant outcomes rather than putting the suitable programs into action. (Akwara, A. F., 2013; Ezirim, Udaw, & Akwara, N. F., 2013). Many parties did not have an appropriate procedure to analyse either the unpredictable market or their opponent. Market analysis ensured that parties would get the maximum attention, especially in each constituency. The quality of the political product: Candidates, ideology or policy were determined by the voters insight directly. Accordingly, the real missing concept in contemporary political marketing was political product development. Interaction with society and members and support by political parties were not always successful. Inappropriate materials (context) and tools (interface) may have caused failure. Mark Triffitt (2016) for example conceded that a significant mistake was 75

96 usually made when parties did not seem to stay consistent with voters interests and did not manipulate the best-fit tools for successful branding. Many parties ignored media and campaign power (Juholin, 2001). In particular, communication via computer networks and new technology opened new opportunities for parties to communicate and promote categories of messages. When these factors were not realised, the online social network was unable to fully utilise this opportunity to connect with potential and actual supporters. Poor planning and management of campaigns was another pitfall, especially in terms of campaign design through the message, identity, length, budget and campaign theme. Did not pay much attention to the significance of conducting an appropriate campaign and promotional activities to build a desirable public image. A poor public image had been reported to arise from several political marketing miscalculations and mishandling. Lock and Harris (1996) asserted that political marketing was like notfor-profit marketing, which was based on trust; It is highly dependent on public image and is particularly at risk of inappropriate representatives, members, policies, and advertisements that have the potential to the mishap the image conveyed. Westling (2007) concluded from his study that Facebook users choose political messages to view whether their own views are represented. Facebook users can take action and create a new group or organisation and encourage others to join. Esposito (2012) described the features of Facebook and Facebook pages that led to the operationalisation of participation into two classes: Active participation and passive participation. He found that passive online participation did not require much participation or interaction from the user - no interactivity and the social aspect of online participation was related to the concepts of the Web. 76

97 Some examples of passive online participation are: Reading posts, notes, or other written information, viewing photos or watching videos. Active online participation, on the contrary, is more interactive and social. Some typical behaviours in this category are commenting on posts, sharing posts written on Facebook pages or even composing posts to support the candidate that appears either on a Facebook wall or on their personal profile. Passive and active online participation are not exclusive categories of users behaviour on Facebook and other social media; rather they are two independent dimensions for describing users behaviour and online participation. Users may exhibit both active and passive behaviour with varying frequency. Park et al. (2009) studied college students` utilisation of Facebook groups to find out the uses and gratifications of user groups and their impact on civic and political engagement. In Facebook, groups are a feature, separate from pages or user profiles. Illustrating the highest uses and gratifications, their research showed that motivated Facebook group usage should satisfy the society and entertain needs. Their research supports previous studies although they did not use the same category names for uses and gratifications. They further found there was a stronger correlation with offline civic and political engagement when the uses and gratifications for using Facebook groups were for informational purposes versus recreational ones or meeting new people. 2.8 Conceptual Development of the Political Marketing Model. This section configures the detailed interactions of the proposed conceptual framework as illustrated below based on the success and/or failure of previous political marketing models. The present study employed Internet utilisation, candidate profiles, party orientation, political aptitude, and voters Facebook likes as predictors of their decision (Figure 2.4b). 77

98 Figure 2.4b Research Journey on the Voters Behaviour Model Vrontis and Thrassou (2007) introduced a framework for political parties, which have similarities to a new business-customer relationship in developed countries. According to this model, customers, treated in our research as voters, move from one point to another asking for their needs and wants, just like that of voters (Figure 2.4b). On the other side, businesses, or candidates, try their best to fulfil these demands to gain profit. In the political context, this would be like voters expressing their demands to candidates, and candidates trying to meet voters will vote for them (Figure 2.5). 78

99 Figure 2.5 Framework for Political Parties, Source: Vrontis and Thrassou (2007) Also adopted from Thrassou et al. (2009), this model can refer to party orientation in this study, and political activities of the candidates and the parties. Party orientation has three dimensions as construct measurements; marketing orientation, product, and sales orientation all of which affect voters actions or, in other words, voters behaviour at the ballot box (Figure 2.6). 79

100 Figure 2.6 Developed Countries Marketing Communications Framework for Small Political Parties. Source: Thrassou et al. (2009) 80

101 Figure 2.7 Individual Political Choice Behaviour, Source: Newman (1999) Newman (1999), in the Individual Political Choice Behavior Model, illustrated the link between candidate personality and the individual s political choice behaviour. According to this model, candidate personality is an aspect that can be assessed from the candidate s profile (Figure 2.7). Therefore, there is a relationship between candidate profile and voter behaviour, according to Newman`s model. Another feature represented in this model is the social imagery of the candidate, which again corresponds to candidate profile. Furthermore, based on this model, political issues are closest to political activities, which have a direct relationship with voter behaviour. As a result, the relationship between candidate profile and political activities are supported by Newman`s model (Figure 2.7). 81

102 Figure 2.8 Ideal Political Marketing Orientation Profile. Source: Ormrod and Henneberg (2011) The Ideal Political Marketing Orientation Profile Model developed by Ormrod and Henneberg (2011) illustrated the dynamics of all the four classes of party orientations, namely the internal orientation, societal orientation, competitor orientation and voter orientation in trying to to achieve the goals of consistent strategy implementation, namely information generation, information dissemination and members participation (Figure 2.8). Political parties also conduct different kinds of political activities, such as getting voters information, to reach the final stage of earning votes. Consistent strategy implementation and voting for candidates have something in common in that they both contribute to the end user, whether a business or a candidate, being able to achieve their goal. On the other hand, as far as voters are concerned, they will also go through three steps to obtain information about candidates they want to vote for, communicate the information among each other and then support the candidate they choose and become a member of their campaign (Figure 2.8). 82

103 Figure 2.9 Marketing Communications Framework for Small Political Parties In the Marketing Communications Framework for Small Parties Model (Figure 2.9), adapted from Thrassou et al. (2009), marketing plan and marketing mix are equivalent to political activity, which affects voter action, or what is termed voter behaviour, in this political marketing context. Therefore, this model supports the relationship between political activities and voting behaviour. 83

104 Figure 2.10 Strategic Marketing Communications Framework for Small Political Parties. Source: Thrassou and Vrontis (2006) Thrassou and Vrontis (2006) in their Strategic Marketing Communication Framework for Small Political Parties suggests it is necessary to collect environmental effects, analyse them and then plan strategically according to the strategy process outcome, which in political marketing corresponds with party orientation activities. Another stage of this model, implementation, falls under the category of political activity, which in turn influences voting behaviour (Figure 2.10) 84

105 Figure 2.11 Business Marketing and Political Marketing Compared. Source: Kotler (1975) The Business Marketing and Political Marketing Comparison Model developed by Kotler (1975) had shown in the business setting section; the buyer pays money to the seller to receive goods and services in return. A similar transaction occurs in political marketing, but the difference is that in the political context, there is no monetary exchange. Thus the voter votes for the candidate in the hope of benefiting from the promises and favours the candidate are expected to provide in future. This exchange is linked to the political activity of the candidate as well as to the political aptitude of the voter, in the way that each party performs its part (Figure 2.11) 85

106 Constituency Ranking Competition Determination Identification of Voting Groups Party Positioning Voter Group Targeting Post Election Analysis Figures 2.12 Model of Political Planning for Local Campaigning Source: Baines et al. (1999c) Baines, Lewis, and Yorke (1999c) suggested a political planning model, which can also explain political activities of candidates (see Figure 2.12). Political parties need to know the level of their voters allegiance to evaluate the likelihood of their success; thus, they need to have historical data of the voters, which they can obtain through constituency research. The data include both party orientation and political activities of candidates and their parties. Party positioning that has been projected in this model was like political activity, and it was described as the process of connecting with the voters. 86

107 Figure 2.13.Theory of Planned Behaviour Source: Hansen & Jensen (2007) The Theory of Planned Behavior Model developed by Hansen & Jensen (2007) as shown in Figure 2.13 supports the notion of voters attitudes towards the candidate, hence determining the voting decision, i.e. whether or not to vote for a candidate. Therefore, it can be said that voting behaviour and intention to vote are the product of voters attitudes towards candidates (Figure 2.14). Figure 2.14 Relationship Marketing in Predictive Model in the Political Process, Source: Dedar& Gray (2011) 87

108 The model developed by Dedar and Gray (2011) known as Relationship Marketing in Predictive Model in Political Process (Figure 2.14) also suggests that voters will not hesitate about voting for a candidate when they believe that he or she is trustworthy. Candidates relationships with voters, especially on a one-to-one basis via social media, have also been shown to be significant in gaining voters allegiance, thus improving predictions of voting behaviour (Figure 2.15). Therefore, there is a correlation between what has been argued here in this research and the model presented (Mahmud & Gray, 2011). Figure 2.15 Channels of Party Communication, Source: Norris (1997a) 88

109 The Channels of Party Communication Model by Norris (1997a) suggests that in the virtual world and social structure, the Internet, especially through social media, play a threefold role. Voters get information about candidates from the candidate profile shown in social media like Facebook. They become persuaded whether or not to vote for the candidates and they will eventually move toward or away from the candidates they have investigated (Figure 2.15). Figure 2.16 Political Marketing Process. The model of Political Marketing Process adapted from Niffenegger (1998) indicates that candidates and parties market themselves by projecting the candidates profile using the 4P s marketing mix Candidates profile in a political marketing context is similar to product image in a business setting as both functions in a similar manner. When candidates make promises, they are in a sense advertising themselves; just as what happens in product promotion in a business context (Figure 2.16). 89

110 Figure 2.17 Buyer Decision Processes, Source: Kotler and Armstrong (1990) In the model (Figure 2.17) known as Buyer Decision Processes developed by Kotler and Armstrong (1990), problem recognition is like that of voters wants and needs or requirement in a political marketing context. Information search is equal to candidate search and evaluation of alternatives is similar to candidate selection. All these steps would bring about purchasing decision or, in the political context, a voting decision and consequently can be considered to be voting behaviour. Evaluation of other alternatives can correlate to examining other candidates and finally deciding whom to vote for (Figure 2.17). 90

111 Figure 2.18 Candidate Marketing Map The candidate profile in the Candidate Marketing Map adapted from Kotler and Kotler (1999) as shown in Figure 2.18, is where candidates are assessed by the voters. Candidates and their parties come up with strategic planning and goal setting through political activities. They communicate with voters using social media, and the outcome of all these efforts will have an impact on voters behaviour, thus earning votes from them. 91

112 Figure 2.19 Newman s Model of Political Marketing, Source: Newman (1994) The Newman (1994) Model of Political Marketing as shown in Figure 2.19 pointed out market (voter) segmentation under which appeared assess voter needs, which in our setting is achievable by carrying out of political activities by candidates. The next items profile voters and identify voter segments were also categorised under political activities in this research. Candidates tried to market themselves and demonstrated their characteristics using what, in this research, is termed Candidate Profile. This concept is also backed by Newman s (1994) model under the category candidate positioning. 92

113 Figure 2.20 Model of Political Communication This Model of Political Communication based on Negrine (1994) recognises the media effects of cognition and attitudes of voters which impact their voting behaviour. It supports the notion that Internet through social media does affect on voters and the way they decide to vote for candidates. This is particularly true for Facebook (Figure 2.20). 93

114 Figure 2.21 The Alternative Model of Political Marketing, Source: Newman (1994) The Alternative Model of Political Marketing by Newman (1994) encompasses all the independent variables identified in this study. Party orientation strategies take care of competitors. Messages that are disseminated using social networks support Internet utilisation in our research. Policy implementation and mutual trust building point to political activities. Permanent communication involves using the Internet and thus checking candidate profile, and finally, lobbying is a kind of political aptitude (Figure 2.21). 94

115 Figure 2.22 Strategic Process for Segmenting Political Markets: The STP Approach. Source: Smith and Hirst (2001) Smith and Hirst s (2001) model of Strategic Process for Segmenting Political Markets: the STP approach supported the independent variables of party orientation and political activity. Segmentation refers to the policies and attitudes to which parties are inclined, while target selection and developing appropriate marketing mix are types of political activities (Figure 2.22). Figure 2.23 Political Marketing Model Finally, the Political Marketing Model developed by Marek (2001) illustrated that voters choose their candidates based on the information they see on the candidate profile (Figure 2.23). As a result, this model correlates with candidate profile, party orientation and political activity, all of which are expanded in the current study. 95

116 Figure 2.24 Navigation System and Functionality of the Reserved Area of the Site This model best describes Internet utilisation in a political marketing context (Figure 2.24). Activities like finding friends and events, fundraising, and sending/receiving messages, are carried out through utilising the Internet and social media, i.e. Facebook. Figure 2.25 Evolution of the Brand Construction Process Candidate profile is a means through which a candidate is branded and marketed. Once this has occurred, it results in customer (voter) commitment. Earning voters commitment means affecting their voting behaviour in a desirable way and therefore benefiting from their votes (Figure 2.25). 96

117 2.9 Summary In psychology and social studies, Exchange Theory has commonly been defined as a perspective, which clarifies social change and constancy in the process of negotiated exchanges between the parties concerned. Previous literature stressed that Exchange Theory often views social order as the unplanned outcome of the exchange between members of society. Cherry (2010) concluded that social behaviour is the impact of an exchange process. Its purpose is to maximise the benefits and minimise the costs. The theory explained how parties in an exchange would weigh the potential benefits and risks of social relationships. The parties naturally discontinue a relationship when risks outweigh the rewards. The exchange theory can also dwell further on the role of social psychological and sociological perspectives in social change and stability as far as negotiating exchanges between parties are concerned. Through the use of a subjective cost-benefit analysis and a comparison of alternatives, the theory, which was actually rooted in economics and sociology elaborates on the human relationships established. Our present study has the aim of ascertaining whether Exchange Theory can be applied to political marketing strategies and whether or not the social media, particularly Facebook can serve as an effective media for the applications of political marketing strategies. From a review of the previous literature, it seems likely that this study is the first research in Malaysia on political marketing strategies in relation to Facebook interactions that have used quantitative methods with a structural equation modelling approach (Figure 3.1b and Figure 3.1c). 97

118 The study is also the first empirical research that specifically targets the younger voter group in Malaysia through an analysis of political interactions on Facebook (Figure 3.1c). There have been numerous discussions on the theory of political marketing, which has been pioneered by Kotler (1975). However, until today, there was no effort made to study the significant role of political marketing in Malaysian politics in this era of the Internet and social media. The underlying theory initially referred to was the traditional economic model, sometimes termed Exchange Theory. This theory can be traced to the concept of generalised exchange posited by anthropologist Bronislaw Malinowski in 1922, who emphasised rewards and costs underlying social interactions (Ceshire, 2013). The Economic Model assumes that things are exchanged for their economic or utilitarian value, and the Economic Exchange Theory model has been implied in the economic discipline since its inception. Price Theory and the Rational Choice Theory, for example, were all based on this reciprocity assumption. Exchange Theory focused on reciprocity, which implies the existence of balancing forces that promote equilibrium (Blau, 1964). However, the simultaneous existence of exchange and power forces produces an imbalance in social life. Initiated by Homans (1958), social exchange theory grew out of the intersection of economics, psychology and sociology. It was developed in an attempt to appreciate the human social behaviour in economic undertakings. It is fundamentally different from Economic Exchange Theory from the perspectives of the actors. Economic Exchange Theory views actors (person or firm) as dealing not with another actor but with a responding market on various market characteristics. The Social Exchange Theory, by contrast, sees the exchange relationship between specific actors as actions contingent on the reward reactions received from other people (Blau, 1964). Homan s work has effectively split Exchange Theory into 98

119 two main theoretical strands namely the Economic Exchange Theory and the Social Exchange Theory. 99

120 CHAPTER 3 RESEARCH METHODOLOGY 3.1 Introduction This study attempted to examine the factors affecting the voting decisions of Facebook users as potential mainstream voters in the Malaysian general elections (see Figure 3.1a). Hence, this study hoped to examine the issue of political marketing in Malaysia and the moderating effect (particularly) the push factors of Facebook Time and Facebook Friends on voters behaviour (Figure 3.1a). Based on Exchange Theory as applied to political marketing, namely Exchange Theory in Political Marketing (PMET), the study examined what major factors are likely to be taken into account by young voters in Malaysia when they vote. The applicability of PMET presumes that voters give their votes with the expectation that they will receive their consideration in the foreseeable future. Overall, the promises will be related to the political candidates, the party they are representing and the circumstances and nature of the promises. The moderating effect of Facebook interactions was evaluated. 3.2 Research Design This study was conducted by means of descriptive research, examining the relationships among variables and testing the validity of hypotheses. Descriptive research entailed the observation and description of variables as they were disseminated across a population (Tabachnick & Fidel, 2007). Broadly speaking, descriptive research designs are determined by written or spoken words and/or observable behaviours serving as the data sources (Hair, Black, Babin, & Anderson, 2009). Quantitative descriptive design produces numeric or statistical descriptive data concerning the manner in which variables are circulated among members of a population (Sekaran & Bougie, 2010). 100

121 The present study employed a descriptive research mode to configure interaction between predictors and voting behaviour and also investigated a moderating influence of Facebook time and Facebook users in the hypothesised model (Figure 3.1a). Based on the underlying PMET and review of the predictors on the voter behaviour, the hypothesised model was established (Figure 3.1b). Using quantitative approach and structural equation modelling (SEM), the study established a voters behaviour model referring to the 12 th General Election (GE12) and validated based on the 13 th. General Election (GE13) Data. Figure 3.1a configured the model journey, research design and also model validation process. Figure 3.1a Research Design and Model Journey 101

122 3.3 Theoretical Framework Following the extensive review of the literature, the present study hypothesised a structural model of voting decisions (Figure 3.1b). There were five predictors serving as the exogenous variables and the voting decision being the endogenous variable. Internet utilisation, candidates profiles, party orientation and political aptitude were hypothesised as predictors of voting decision. Furthermore, the present study also hypothesised Facebook time and Facebook friends as moderating variables on the relationship between predictors and voting behaviour (see Figure 3.1c). Internet Utility Internet Utilisation Product Candidate Profiles Sales Party Orientation Voting Decision Marketing Political Aptitude Political Activities Figure 3.1b Hypothesised Model 102

123 Internet Utility Internet Utilisation Product Candidate Profiles FB Time Sales Party Orientation Voting Decision Marketing Political Aptitude FB Friend Political Activities Figure 3.1c Hypothesised Model with Moderating Variables (FB Time and FB Friend) 3.4 Research Hypothesis Based on the hypothesised model (see Figure 3.1b), the present study also hypothesised 17 statements as a causal relationship examination between exogenous and endogenous variables. Hypothesis 3 and Hypothesis 4 tested the moderating variables in the structural model (Figure 3.1c). Following are the hypotheses for this research study. In essence, there were 5 main hypotheses in the research. However, there were 5 subs hypothesis for hypothesis 1, five sub-hypothesis in hypothesis 3 and five sub-hypothesis in hypothesis 4. The sub-hypothesis was to probe for further details the role of internet utilisation, candidates profile, party orientation, political orientation, political aptitude and political activity. 103

124 H1 There is a direct effect of Predictor on the Significant predictor on the Voting Decision as the Dependent variable. 1. H1a Internet Utilisation has a direct and significant influence on Voting Decision 2. H1b Candidate Profiles have a direct and significant influence on Voting Decision 3. H1c Party Orientation has a direct and significant influence on Voting Decision 4. H1d Political Aptitude has a direct and significant influence on Voting Decision 5. H1e Political Activities have a direct and significant influence on Voting Decision 6. H2 Product, Sales and Marketing confirm the construct of Party Orientation There is a moderating effect of Facebook Likes and Facebook Utilisation 7. H3a Facebook Likes has a moderating influence on the relationship between Internet Utilisation and Voting Decision 8. H3b Facebook Time has a moderating influence on the relationship between Candidate Profiles and Voting Decision 9. H3c Facebook Time has a moderating influence on the relationship between Party Orientation and Voting Decision 10. H3d Facebook Time has a moderating influence on the relationship between Political Aptitude and Voting Decision 11. H3e Facebook Time has a moderating influence on the relationship between Political Activities and Voting Decision 12. H4a Facebook Friends has a moderating influence on the relationship between Internet Utilisation and Voting Decision 13. H4b Facebook Friends has a moderating influence on the relationship between Candidate Profiles and Voting Decision 14. H4c Facebook Friends has a moderating influence on the relationship between Party Orientation and Voting Decision 15. H4d Facebook Friends has a moderating influence on the relationship between Political Aptitude and Voting Decision 104

125 16. H4e Facebook Friends has a moderating influence on the relationship between Political Activities and Voting Decision 17. H5 Internet Utilisation, Candidate Profiles, Party Orientation, Political Aptitude, Political Activities, moderated by Facebook Time and Facebook Friends confirm the significant interactions in the Structural Model. 3.5 Variables and Measurements A Likert scale was used for sections 2, 3 and 4 of the questionnaire (see Appendix 1). The Likert method is based on the assumption that an overall score generated from responses to the items reflecting a particular variable provides a reasonably good measure of the variable. Likert scales are often used to measure opinions or attitudes of individuals. Five categories namely: - Strongly agree, - Agree, - Neutral, - Disagree, and - Strongly disagree. Experts recommend using an equal number of positive statements for each item being measured (Hair et al., 2009). 3.6 Population and Samples One of the analytical techniques used in this study is structural equation modelling. A few researchers suggested that for SEM, sample sizes should range from 150 to 250 to prevent misspecification from occurring. Typically, a sample size of at least 150 would provide parameter estimates that have standard errors which are small enough to be of practical usefulness (Anderson & Gerbing, 1988). Generally speaking, a sample size of 200 would be the best number that makes the critical sample size (Byrne, 2001). 105

126 Hair et al. (2009) conceded that sample size should be increased under the following conditions: - Misspecification probable; - The model is too complicated; - The data show several non-normal characteristics; or - An alternative estimation procedure is adopted. Bentler (1993) suggested using a sample size of at least 5 cases per parameter to be appraised in the structural equation modeling. Hoelter (1983) addressed the issue of sample size sensitivity by providing a formula for what he termed the critical-n (CN). This CN is the size that a sample must reach in order to accept the fit of a given model on a statistical basis. The present study recruited 505 respondents, a number which is well above the acceptable critical sample size of guaranteeing that the respondent s size is credible. 3.7 Data Collection Technique Questionnaire Design The primary data collection for this research was via a self-response questionnaire. The use of questionnaires as tools of collection for primary data in academic research is a controversial issue. Much has been written concerning questionnaires as research instruments for gathering information from people. According to Sekaran and Bougie (2010), a good questionnaire must be free of ambiguity; it must be precise, clear, adequate; and every single question should express a single idea in order to fulfil the questionnaire s purposes. Tabachnick and Fidell (2007, p.62) argued that when using a questionnaire, the researcher should attempt to capture the values, perceptions and interests of the respondents and they defined questionnaires as research tools through which people are asked to respond to the 106

127 same set of questions in a predetermined order. Sekaran and Bougie (2010) also state that in questionnaires, respondents read the questions, interpret what is expected and then write down the answers. They consequently argued that in the case of a questionnaire, as there is no one to explain the meaning of questions to respondents, it is important that questions be clear and easy to understand. According to them, this is in contrast to the position in interviews where the interviewer can ask, explain the questions and then recode. Questionnaires are fundamental elements that can address the needs of researchers but not a useful mechanism for assessing respondents. Burgess (2001) insisted that researchers need to review the literature and, in some cases, do preliminary research on their target population. In this way, researchers are able to design (and re-design) their questionnaires and select questions that are relevant, concise and efficient. In other words, one can conclude that the content, structure, format and sequence of questionnaires must be clear and understandable in order to motivate the respondents to provide the information required. The questionnaire was designed to assess practical attitudes and beliefs. Therefore, before the final questionnaire was produced and distributed, a pre-test and pilot study was performed in three stages and carried out by selecting respondents at random (Sekaran & Bougie 2010; Tabachnick & Fidell 2007). The pilot study at every stage was carried out just like in the main study. At every stage, the questionnaire was sent out to small groups of people consisting of between respondents within Kuala Lumpur. The purpose of conducting the pilot study in three stages was to reveal any weaknesses or contradictions in the questions and to correct them prior to the main survey while at the same time testing and improving reliability and validity where the questions posed elements of 107

128 vagueness and difficulty. It was also essential to assess the length of time respondents might need to complete the questionnaire. The researcher had two purposes in mind: To maximise the response rate and to minimise error frequency (Hair et al., 2009). Finally, it was important to test the questionnaire's ability to obtain the type of data anticipated. The pilot test indicated that changes needed to be made before the final draft of the questionnaire could be completed and distributed. In addition, positive reactions from the pilot study were also taken into account when constructing the final questionnaire. In most cases, researchers cannot make direct observations of every individual in the population they are studying. The term population or study population refers to all members of the group or society that researchers are interested in. Sampling is the process of selecting a group from the population to form a basis for estimating, studying and predicting facts, attitudes or outcomes regarding the population in general. This selected group is called the sample: The sample is a subset of the population that is usually chosen to include and represent all members of the population to achieve the purpose of the study (Tabachnick & Fidell, 2007) Sampling Mechanism The number of registered voters in 2008 stood at around 10.1 million as indicated by the official statistics of the Malaysian Election Commission. Datuk Wan Ahmad bin Wan Omar, Deputy Chairman of Malaysian Election Communication (personal communication, March 18 th, 2012) states that 50% of registered voters in 2008 are young voters and that young voters significantly affected the outcome of election as they formed 80% of the newly registered first-time voters. 108

129 The present study adopted a Stratified Random Sampling mechanism to determine and identify respondents for this study s survey (Tabachnick and Fidell, 2014; Hair el at., 2013; Sekaran & Boogie, 2013). Sets of questionnaires using a five-point Likert scale were distributed to 784 samples as suggested by Krejie and Morgan (1970) to minimise the margin of error. The Malaysian Center for Public Opinion s (MCPO) database of eligible voters was used as sampling frame. This database contains such information as national ID number, date of birth and home address. The home address was used to reach out to each of the research respondents. A young voters category of ages between 21 to 30 years old were involved in the survey (Table 4.1) as they formed more than 50% of total 800,000 Facebook users in 2008 (Source: Allfacebook.com/Facebook-demographic-statistics as of 16 th December 2008). The 784 sample questionnaires were distributed and 550 questionnaires were returned (70.1% response rate). After outliers were identified, 505 surveys (64.4% response rate) were accepted for data analysis, model establishment and hypothesis testing (Hair et al., 2013) Validation of Data Gathering The fairness of any assessment tool firstly depends on the procedures followed in its standardisation, validation and reliability (Erasmus, 1997). Researchers in the social sciences naturally look into various complex constructs where reliable and valid measures would be vital. It is important for the measures to be concise, clear, and easy to be carried out (Rubio, Berg-Weger, Tebb, Lee, & Rauch, 2003). Psychometric testing is needed to measure the variables. First of all, the reliability and validity of the psychometric test, for that specific situation have to be determined. 109

130 Smith and Robertson (1986) defined reliability as the extent to which scores on a measure escape the random errors. Price (1997) explained that reliability is regarded as the extent to which a measure generates the same results when used over and over again. Test reliability thus refers to the instrument's degree of accuracy and consistency. Reliability indicates the measure of confidence with which scores obtained with an instrument can be regarded. According to Schaap (2003), one cannot refer to the reliability of an instrument; instead, reliability simply tells one how confident one can be that the scores obtained with the instrument are consistent and accurate. Reliability is relative and can be influenced by the group to whom the instrument is administered. Reliability, when used in the appropriate formula, will indicate the relation between true scores and observed scores (Ferdinand, 2002). This is termed the standard error of measurement, which provides a direct indication of the degree of inconsistency or error one can expect with individual scores (Byrne, 2001). According to Byrne, there is no fixed value below which reliability is unacceptable and above which is acceptable: it depends on what the purpose of the score is. Cognitive-related ability instruments should have higher reliability coefficients. According to Hair et al. (2009), Cronbach s alpha measures the intercorrelation of items. If alpha is greater than or is equal to 0.60, the items are regarded as unidimensional, and they may be integrated into an index or scale. In this work, the more stringent cut-off of 0.70 had been used. Cronbach s alpha is the most common form of internal consistency reliability coefficient (Kelly, Downey & Rietdijk, 2010). Alpha equals zero when the true score is not being measured at all, with only an error component, while alpha equals 1.0 when all items only measure the true score, and there is zero error component. 110

131 Cronbach s alpha refers to the percentage of variance the observed scale would justify in a hypothetical true scale composed of all possible items in the universe. It can be interpreted as the correlation of the observed scale with all possible other scales that weigh on the same thing and using the same number of items. An indicator can be reliable without having to be valid (Smith, 1991). For it to be valid, however, it has to be reliable (Smith,1991). Thus, reliability can be seen as necessary, but the requirement for validity can be inadequate. Just as reliability does not refer to a single attribute of a test, the concept of validity is also multifaceted. 3.8 Data Analysis Procedure Structural Equation Modeling (SEM) The analytical approach entailed processes such as scale validation, scale dimensionality, exploratory and confirmatory factor analyses to ensure that the measures were psychometrically thorough (Arbuckle, 2006). The scale validation and dimensionality indicated that a preliminary examination of reliability and validity of the measures was done using Cronbach s alpha and confirmatory factor analysis (Ghazali, 2003). In addition, convergent validity, discriminate validity, nomological validity of exogenous and endogenous constructs combination of exploratory and confirmatory factor analyses were examined. The conceptual model and the associated hypotheses were tested with the help of the Structural Equation Modeling ( Anderson and Gerbing, 1988). Analysis of moment structure (AMOS version 21.00) was used in the confirmatory factor analysis. The maximum likelihood estimation (MLE) method was adopted (Joreskog & Sorbom, 1984). The MLE was used for estimating the model parameters with the covariance matrix as data input. Moreover, this method had been described as being well suited to theory testing and development (Bollen, 1990). 111

132 Using SPSS and structural equation model (AMOS 21.0) for the confirmatory factor analysis (CFA) and path analysis; the results were analysed to confirm the validity of the scales and structural relationships among all the independent, moderating and endogenous variables. The present study created a model to test relationships among the observed and latent (hidden) variables using the SEM, and then it was used to check on the hypotheses and verify the relationships established. (Kenny, 2008). Structural equation modeling (SEM) covers many statistical techniques. The path analysis and confirmatory factor analysis are two examples. In this research, the SEM analyses were carried out using AMOS 21.0 version. (Arbuckle, 2012). SEM has purposes that are similar to those of multiple regression, but it is more powerful, given that it considers the modeling of interactions, nonlinearities, correlated independents, measurement error, correlated error terms, multiple latent independents measured by multiple indicators and one or more latent dependents that also come with various indicators (Byrne, 2001). According to Byrne, SEM functions are a more robust alternative to multiple regression, path analysis, factor analysis, time series analysis and analysis of covariance. This suggested that the procedures may be regarded as special cases of SEM, or, to put it another way, that the SEM is of the follow-up of the general linear model (GLM) of which multiple regression is a part (Bollen, 1989). SEM software can test not only the traditional models but also the more intricate relationships and models (Bollen, 1989). As compared to multiple regression, the SEM boasts a few plus points including more lenient assumptions (particularly allowing interpretation even in the face of multicollinearity), the adoption of the confirmatory factor analysis to minimise the measurement error by having multiple indicators per latent variable, the appeal of SEM s graphical modeling interface, the favoured testing models (rather than individual coefficients), the ability to test models with 112

133 various dependents, the ability to model mediating variables rather than be limited to an additive model, the ability to illustrate error terms, the ability to examine the coefficients across multiple between-subjects groups, and the ability to administer complicated data (Tabachnick & Fidell, 2007). There are two kinds of variables that can be included in the SEM: Observed variables and latent variables. Observed variables carry some data and the numeric responses to a rating scale item on a questionnaire. Observed variables in SEMs are on-going, by nature (Bollen, 1989). Latent variables are not observed in a direct manner but only when researcher still wishes to know further about them. To observe them, a model was built and tested to ascertain the latent variables in terms of observed variables. SEM refers to an extension of the general linear model (GLM) allowing a researcher to test a set of regression equations in one go (Hair et al., 2009). The first step was to specify a model depending on a theory and then to see how to measure the constructs, gather the data, and then input the data into the SEM software package. The package adjusted the data to the specified model, so they fit and yielded the results, which include the overall model fit statistics and parameter estimates (Sentosa & Nik Kamariah, 2012). Normally, the input to the analysis is a covariance matrix of measured variables such as survey item scores, although, at times, the matrices of correlations or matrices of covariance and means are used (Byrne, 2001). Practically, the data analyst usually provides SEM programs with raw data and the programs transform these data into covariance and means for its own purpose. The fundamental approach to performing an SEM analysis is shown below in Figure 3.2 (Hair et al., 2009). 113

134 Figure 3.2 SEM Procedure The model shows a set of relationships among the measured variables, expressed as restrictions on the total set of possible relationships (Figure 3.2). Figure 3.3 Flowchart of Covariance of the Structural Model The results also exhibited the overall indexes of model fit and the parameter estimates, standard errors, and test statistics for each free parameter in the model (Figure 3.3). As noted by Tabachnick and Fidel (2007), there are several attractive virtues of the SEM: 114

135 Assumptions made for the statistical analyses are clear and testable, giving the investigator full authority and potentially improving one s understanding of the analyses. Graphical interface software does a lot of creativity, and it helps boost the rapid model debugging (a feature limited to selected SEM software packages). SEM programs simultaneously provide overall tests of model fit and individual parameter estimate tests. Regression coefficients, means, and variances may be drawn in comparison in one go, even across various between-subjects groups. Measurement and confirmatory factor analysis models can be used to purge errors, lessening the contamination of the estimated relationships among latent variables by the measurement error. Ability to fit non-standard models including lenient handling of longitudinal data, databases with autocorrelated error structures (time series analysis) and databases with non-normally distributed variables and partial data. Finally, SEM offers a unifying framework under which numerous linear models may be fit using software that is flexible and powerful (Kabanda, 2013). Being an easy-to-use structural equation Modeling (SEM) program, AMOS examines the relations among observed and latent variables and then adopts those models to check on the hypotheses and endorse the relationships Goodness of Fit Indexes Model fit decides on the degree to which the structural equation model matches the sample data. The determination of model fit in SEM is not as straightforward as in other statistical approaches (Schumacker & Lomax, 1996). Goodness to the model fit shows how well is the 115

136 reproduction of the observed covariance matrix by the specified model among the indicator items (Hair et al., 2009). Values of fit indices indicate only the average or overall fit of a model. As a single index reflects only a particular aspect of the model fit, a favourable value of that index does not itself indicate a good fit. SEM fit indices carry no single statistical test of significance that recognises a correct model provided in the sample data (Schumacker & Lomax 1996). This is also why model fit is usually assessed based on the values of more than one index. In other words, there is no single standard index that can give a standard for all models established (Kline, 2005). However, Hair et al. (2009) asserted that the most fundamental absolute fit index is the chi-square statistic (χ2). It is the only statistically based SEM fit measure. Kline (1998) suggested on reporting at least four tests namely chi-square; Goodness-of-fit Index (GFI), Normed Fit Index (NFI) or Comparative Fit Index (CFI); Non-Normed Fit Index (NNFI); and Standardised Root Mean Residual (SRMR). Schumaker and Lomax (2004) also proposed chi-square with degrees of freedom (DF) and P level, normed fit index (NFI), GFI, CFI and Root Mean Square Error of Approximation (RMSEA). Apart from the assessment of fit statistics to recognise a good fit between the data and the model, one good recommendation is that the individual item reliability and internal consistency of the measurement scale should be studied (Byrne, 2001). The assessment of the individual item reliability can be gauged by studying the indicator loadings and the t-value related to each loading to identify the statistical significance. Hair et al. (2009) conceded that the t-value of each factor loading should surpass the critical value of 1.96 for a two-tailed test and for a one-tailed test at 0.05 significance level to be statistically significant for the construct proposed. In addition, the magnitude of standardised regression weight or factor 116

137 loading should be equal to, or greater than 0.5 for adequate individual item reliability, together with supporting the convergent validity (Bagozzi & Yi, 1989). Significantly high factor loading is an indicator verifying the relationship between indicators and constructs. In addition, Squared Multiple Correlation (SMC) for each variable demonstrate the amount of variance in the respective measures because it is another indicator to be analysed. Byrne (2001) conceded that the Squared Multiple Correlation is to be equal to or larger than 0.5, indicating the extent to which the variable concerned serves as an indicator of a latent construct. Thus, items with non-significant t-values and SMC not more than 0.5 may be removed from the model modification. This item-deletion procedure has been suggested as a method for re-specifying indicators that do not work out (Ferdinand, 2002). Fornell and Larcker (1981) re-emphasised on the importance of studying the composite reliability and variance extracted measures for each construct in the evaluation of the measurement scale s internal consistency. Another complementary measure of the construct reliability value is the variance extracted measure (Tabachnick & Fidell, 2007). This measure shows the overall amount of variance in the indicators considered by the latent construct. Higher variance extracted values occur when the indicators genuinely depict the latent construct (Ferdinand, 2002). In general, the value of variance extracted for a construct should be greater than or equal to 0.50 (Fornell & Larcker, 1981) possibly meaning that the variance weighed upon by each construct is greater than the variance accounted for by the measurement error. A model modification is used to improve the model fit through adding or deleting estimated parameters from the original model (Nik Kamariah & Sentosa, 2008). The empirical indicators of model modification come from an examination of the residuals of predicted 117

138 covariance or the correlation matrix. Residual values greater than ± 2.58 are considered to be statistically significant at the 0.05 level, thereby indicating a substantial prediction error for a pair of indicators according to (Tabachnick & Fidell, 2007). Another indicator of model modification is the modification index (MI) related to the expected parameter change statistics (EPC), which corresponds to the reduction of chi-square simply to obtain a better fit of the data (Hair, Black, Babin, Anderson, & Tatham, 2006). Bollen (1989) also suggested that a fixed parameter with a large MI and EPC should be freed, which would make substantive sense in improving the overall data-model fit. If chi-square is large compared to the standard, the parameter with the largest modification index should be unperturbed if this parameter can be interpreted substantively to enhance the data-model fit (Jöreskog&Lopez,1999). Nonetheless, all post-hoc changes and the modified model should be examined again after the estimated parameters have been modified (Ferdinand, 2002). After the validation of the measurement model, the model fit must be assessed to identify how far the specified indicators depict the hypothesised constructs in the confirmatory factor analysis and the path coefficients of hypothesised relationships between theoretical constructs in the structural equation model (Nik Kamariah & Sentosa, 2008). Broadly speaking, goodness-of-fit is regarded as a measure of the correspondence of the actual or observed input (correlation or covariance) matrix with that predicted by the model suggested. It was argued by Byrne (2001) that the widely accepted condition is that a chi-square to a degree of freedom ratio in the range of 2 to 1, or 3 to 1 indicates an acceptable fit between the hypothetical model and the sample data. The sensitivity of the chi-square statistics to the sample size and model complexity is widely acknowledged; therefore, on this premise, the rejection of a model alone would not be 118

139 appropriate (Bagozzi & Yi, 1989). Accordingly, other measures of fit that compensate for sample size also have to be evaluated on how well the confirmatory factor analysis model reproduces the variables covariance matrix observed. In effect, other goodness-of-fit measures have been recommended to ascertain the model fit, other than an insignificant chisquare. To look at it more specifically, other goodness-of-fit measures can be represented by absolute, incremental, and parsimonious fit measures (Hair et al., 2009). The absolute fit measures firstly examine only the overall model fit with zero adjustments for the degree of overfitting that possibly occurs. The incremental fit measures, next compare the proposed model to a comparison established between models with varying estimated coefficients aiming to find out about the amount of fit achieved by each estimated coefficient (Nik Kamariah & Sentosa, 2008). Table 3.1 exhibits the Goodness-of-Fit (GOF) criteria and acceptable fit interpretation. Overall model fit measures include the Goodness-of-fit index (GFI) (Jöreskog & Yang, 1996). It is a descriptive overall absolute fit measure, which is not explicitly dependent on the sample size and measures how much better the model fits if it is to be compared with no model at all. A minimum value of 0.9 for these two measures is usually considered acceptable, and unity shows a perfect fit (Bagozzi & Yi, 1989) (Table 3.1). The goodness-of-fit test criteria are illustrated in the table below. The same threshold value is applicable to the Adjusted Goodness-fit-index (AGFI), Comparative fit index (CFI), the Tucker-Lewis index (TLI) and the Normed fit index (NFI); and four incremental fit indexes put forth by Bentler (1990). The Root Mean Square Residual (RMSR), Root Mean Square Error of Approximation (RMSEA) and measures of discrepancy per degree of freedom are fit measures leaning on the notion of non-centrality (Bollen, 1989). In normal circumstances, values up to 0.08 can well indicate a reasonable model fit and 119

140 represent reasonable errors of approximation in the population, whereas for the values of about 0.05 or less, they indicate a close fit related to the degree of freedom (Byrne, 2001). Table 3.1 Model Fit Indexes Identification Source: Hair et al. (2009) Model FIT Index Measure Acceptable Level Descriptions Absolute Fit Measures Chi-square statistic Goodness of- Fit Index (GFI) Root Mean Square Residual (RMSR or RMR) Root Mean Square Error of Approximation (RMSEA) Incremental Fit Measures Adjusted Goodness-of-fit Index (AGFI) Tucker - Lewis Index (GFI) Normed Fit Index (NFI) Comparative Fit Index (CFI) Incremental Fit Index (IFI) Parsimonious Fit Measures Normed chi-square (CMIN/df) Akaike s Information Criterion (AIC) Probability Level (P) > 0.00 Value close to or > 0.9 Value < 0.08 Values up to 0.08 are reasonable < 0.05 a very close fit Value close to or > 0.9 Value close to or > 0.9 Value close to or > 0.9 Value close to or > 0.9 Value close to or > 0.9 Ratio between 1 to 2 Smaller, positive values indicate better fit and greater parsimonious. Non-significant chi-square test gives support for the model. 0 (poor fit) to 1 (perfect fit). The closer the RMSR to 0, the better the model fit. RMSEA less than or equal to 0.08 shows a reasonable error of approximation while a value above 0.1 suggests that there is a significant problem. Value adjusted for df, with 0.90 a good model fit. Lower coefficient (closer to 0.50) shows that the relationship amongst variables is more complicated than can be represented by that number of common factor. NFI close to 1 suggests a very good fit. CFI close to 1 suggests a very good fit. IFI close to 1 suggests a very good fit. Wheaton, Muthén, Alwin, and Summers (1977) suggest a ratio of approximately 5 or less as rather reasonable. Carmines and McIver (1981) suggest that the ratio in the range of 2 to 1 or 3 indicates an acceptable fit between the hypothetical model and the sample data. AIC close to 0 indicates an excellent fit. 120

141 3.8.3 Composite Reliability and Average Variance Extracted (AVE) Using the six predictors as latent constructs, a composite reliability test was carried out. Composite scores were computed from the data in multiple variables in order to form reliable and valid measures of latent and theoretical constructs. The variables that had relationships with one another were combined to form a composite score. This was tested through exploratory factor analysis and reliability analysis. Composite reliability is representative of the shared variance among a set of observed variables measuring an underlying construct (Nik Kamariah &Sentosa, 2008). Its potential definition is the proportion of true variances in total variance (true plus error variance). The reliability of an individual measurement can be evaluated via a square multiple correlation (SMC) between the observed score and the true score. Bagozzi and Yi (1989) proposed an acceptable threshold of these two criteria and desirable composite reliability should be greater than or equal to 0.60, with the latest recommendation for it to be greater than or equal to 0.70 (Hair et al., 2009). Average variance extracted (AVE) denotes the average value extracted of two constructs. In this study, the AVE values of the constructs were obtained to discover if the discriminant validity was supported or if multicollinearity was not present (Byrne, 2001). Discriminant validity refers to the extent to which a construct is truly regarded as unique as compared to other constructs (Hair et al., 2009). 121

142 3.8.4 Multicollinearity Test Hair et al. (2009) define multicollinearity as the correlation among three or more independent variables (evidence when one is regressed against the others). As collinearity increases, the unique variance explained by each independent variable decreases. Multicollinearity is a problem with a correlation matrix that occurs when variables are too highly correlated. Multicollinearity causes both logical and statistical problems when redundant variables are included in the analysis, as it inflates the size of error terms and weakens the analysis (Tabachnick & Fidell, 2007). Multicollinearity can have substantive effects not only on the predictive ability of regression model but also on the estimation of the regression coefficients and their statistical tests. As multicollinearity increases, it complicates the interpretation of relationships because it becomes difficult to ascertain the effect of any single construct because of the interrelationships among them (Hair et al., 2009). In this study, multicollinearity was computed using the comparison between average variance extracted value and correlation square among variables (Tabachnick & Fidell, 2007). 122

143 CHAPTER 4 RESEARCH FINDINGS 4.1 Introduction This chapter takes a close look at the findings of the study obtained from the statistical analysis that has been carried out. It starts with the data collection overview and the description of the respondents profile (Hair, Black, Babin, Anderson, & Tatham, 2010). In this chapter, multivariate outliers, normality tests, the reliability and validity of measurements are also discussed. The chapter also presents in detail the findings from the analyses of the data. The findings are systematically presented as responses to the formulated research questions. The findings gathered from the party orientation categories based on the 2 nd order analysis of latent construct measurement have verified the domains of the party orientation categories as exogenous variables established in the hypothesised model (shown in Figure 4.1). Internet Utilisation Figure 4.1. Hypothesised Model using SEM 123

144 The model served to test the hypotheses and to verify the interactional effects among variables in the hypotheses, and the first Re-specified Model. The chapter then proceeds with CFA on the hypothesised model using SEM (Hair et al., 2010; Byrne, 2001). The goodnessof-model fit, direct and indirect effects of variables via the path analysis and correlation among variables are also elaborated. Last but not least, the results of hypotheses testing and also the summary of findings are established. 4.2 Profile of Respondents There were 505 respondents involved in this survey. This section configures detailed profiles of the respondents in the sample. Table 4.1 illustrates the percentages of male and female respondents that participated in the survey: 68.7% of the total numbers of respondents were male and the remaining 31.3% were female. As shown in Table 4.1, the ethnicity of the respondents is categorised into four groups, i.e. Malay, Chinese, Indian and other ethnic races. Most of the respondents in this research (n- 505) were Malay (67.5%), followed by Chinese (25%), Indians (only 5%) and other ethnic groups (2.6 %). In terms of religion, the participants were divided into five groups. Muslim respondents made up the majority with 71.1%; Christians 17.4%; Buddhists 7.9%; Hindus 3.2%; and respondents of other religions 0.4%. Respondents were also categorised according to age. Three categories were used: Namely years, years, and years. Respondents aged years constituted the largest group with 67.7%; those aged years made up 28.5%; while those aged made up only 3.8%. The respondents profile fits well into the population structure by ethnicity. Table 4.1 also shows the locational distribution of the respondents throughout Malaysia. Participants from Kuala Lumpur/Selangor formed the largest group of respondents with 124

145 42.4%. The next largest group was from Melaka with 32.7%. The remaining states had considerably lower percentages as follows: Johor 2.85%, Negeri Sembilan 0.8%, Perak 1.4%, Penang 1.2%, Kedah 1.8%, Kelantan 1.4%, Terengganu 3.6%, Pahang 4.4%, Sabah 4% and Sarawak with 3.8%. All respondents who participated in this research were eligible registered voters. Of all the young people who participated in this research, 65.5% were from urban areas, 25.5% came from rural areas, and 8.95 of the respondents were from other areas. According to the level of education, 62.8% of the respondents held a university degree, 27.3% had a college certificate, and 9.9% held higher secondary school certificate (Table 4.1). Furthermore, in terms of respondents qualifications, most of them had postgraduate qualifications (64.4%), followed by the undergraduate (26.1%) and then diploma holders had a percentage of 9.5. In terms of occupation, more than half of the respondents, i.e. 52.7% were university students, 19.2% were self-employed, 15.8% were business executives, 8.1% were unemployed, and 4.2% were government officers. Most respondents, (81.6%), were single, (15.8%) were married, and 2.6% of them chose not to reveal their marital status (Table 4.1). The participants in the survey were also categorised based on their income status. Those who reported incomes of more than RM 10,000 per month made up 6.3%; 15.6% earned between RM 5,000 and RM10, 000; 1.6% earned between RM 3,000 and RM5, 000; those who earned RM 1,000-3,000 constituted the majority (60.2%), and respondents who earned less than RM 1,000 accounted for 16.2%. As for residence, 63.2% of the respondents reported living in their own houses, and 36.8% were living in rented accommodation (Table 4.1). Furthermore, as far as positions were concerned, the information collected revealed that 45.9% of all respondents were organisation leaders; 27.1% were committee members of organisations; 21.8% were ordinary members of organisations, and finally, 5.1% had not 125

146 joined any organisation at the time of the survey. Leaders constituted 25.9% of the respondents in this research, while followers were 32.7%; those who defined themselves as activists accounted for 34.3%, and those who belonged none of these classes was 7.1% (Table 4.1). Table 4.1 also shows the respondents` political descriptors. As much as 28.9% had joined a political party, 25% had joined and subsequently left a political party, 36.4% said they intended to join a political party, and 9.7% did not have any interest in joining a political party. Of all the respondents, 70.7% were certain about which party they were going to vote for, and 29.3% were undecided voters who had not decided which party they would vote for. The vast majority of respondents participating in the study (93.7%) owned a computer; however, 6.3% did not possess one (Table 4.1). The main concern of this study was the extent of the use of the social media (and Facebook in particular) and its link to Malaysian politics. Although the extent of their use varied, all respondents were Internet users. 31.5% were new users (using the Internet for less than a year); 28.5 % were users with between 1 and 3 years of use; 31.3% had been using the Internet for between 3 and 5 years, and; 8.7% had been using the Internet for 5-10 years were 8.7%. Participants who accessed the Internet from their homes accounted for 10.7%; users who accessed the Internet from their office were 16.6%; those who used free Wi-Fi in different places formed 0.4%; Blackberry internet users constituted 56.4%; and those who accessed the Internet via iphone accounted for 15.8% (see Table 4.1). Participants varied in the type of social media they were using; 46.5% were Facebook users, 19.8% were using MySpace, 8.7% were Friendster subscribers, 11.1% had an account in LinkedIn, another 11.1% were using Twitter, and 2.8% were involved in other kinds of social 126

147 media such as Telegram and Tango. The number of years respondents had been using the social media ranged from less than a year - to more than 3 years; 42.4% stated they had signed up less than a year earlier; 29.9% had been subscribers for 1-3 years, and 27.7% had been using social media for more than 3 years (Table 4.1). Table 4.1 suggests that more than half the sample (63.4%) usually logged in to their social media for less than 30 minutes; 25.5% between 30 minutes to one hour; and 11.1% spent 1 to 3 hours daily surfing their favourite social media. As much as 28.7% of respondents reported using social media all the time; those who were daily visitors formed 19.4%; some 40.4% visited their profiles monthly, and 11.5% chose others. A total of 68.9% of respondents reported that they interacted with politicians; 25.5% reported that they interacted with celebrities, while 5.5% said they interacted with their classmates (Table 4.1) as friend in the social media. A small percentage of participants in this research, 11.1%, had fewer than 50 friends ; 16.6% of them had 51to100friends; more than half of the sample, i.e. 54.7% had friends and those who had more than 500 friends comprised 17.6%. More than two-thirds of the research population that is 71.7% had friends with politicians on Facebook, and the rest 28.3%, had non-politician friends. Respondents interaction preference was as follows: 35.2% were interested in adding politicians as their friends on Facebook, 25.3% wanted to join politicians fan pages, 16.6% preferred to join political parties fan pages, and 22.8% wanted to join political parties Facebook groups (Table 4.1). 127

148 Table 4.1 Profiles of Respondents Variable Indicator F % Gender Male Female Ethnicity Malay Chinese Indian Others Religion Islam Christian Buddha Hindu Others 2 4 Age State of Origin Kuala Lumpur/Selangor Melaka Johor Negeri Sembilan Perak Penang Kedah Kelantan Terengganu Pahang Sabah Sarawak Registered Voters Voters Location Urban Youth Rural Youth Others Education University College Higher Secondary School Qualification Postgraduate Undergraduate Diploma Occupation Self Employed Business Executive Government Officer University Student Unemployed Marital Status Single Married Prefer not to answer

149 Income (RM) More than 10, ,000 10, ,000 5, ,000 3, Less than 1, Residence Ownership Own House Rented House Position in Political Parties Leader in Organisation Committee Member of Organisation Ordinary Member of Organisation Had not joined any organisation Future role in politics A Leader A Follower Activist None of the above Involvement Has joined a political party Has left a political party Will join a political party No desire to join a political party Owning Computer Yes No Using Internet < 1 year years years years Accessing Through Home Office Free WIFI Area Blackberry IPhone Social Media Facebook MySpace Friendster Linkedln Twitter Others Sign up Duration < 1 year years > 3 years Length of Time on the Internet The Frequency of Visit in Social Media < hours All the time Daily Monthly Others Interaction Groups Politician Celebrities Classmates

150 Number of Friends in Social Media < > Type of Facebook Friends Politicians Non Politicians General Political Interaction Preference Add as friend Join politicians' fan page Join political party fan page Join political party Facebook group Messages Posting on the wall Direct message Advertisement Column Other Least preferred political Activity Update about a politician on his daily schedule Politicians argument on current issues Politician update about his family members Politician criticism against other politicians Pre-impression of Politicians Positive Neutral Post Impression More positive Less positive Politician Interaction Political Party Type Interact with me more often on Facebook Make a promise that you can deliver Please improve on handling issues Political party that propagates ideology Political party that promotes services that their party prefers Political party that understands wishes of voters and strives to meet them Invitation To attend their campaign event To attend their family gathering Not interested to meet them in person Knowledge about politics Very Good Good Average Poor Very Poor Political affiliation Those existing in the Cabinet Those existing in parliament Those emerging young politicians Political Activism Will register as voters Have registered as voters Will vote in the coming election Will not vote in next election Level Number of Facebook friends Number of followers on twitter

151 Voting Criteria The candidates good image The candidates future promise The candidates previous favour Image of the leader of their parties Voting Preference I have decided who I will vote for and will not change my mind I have decided who I will vote for but will prepare to change my mind I have not decided who I will vote for Active on Facebook Yes No Usage of Facebook Yes No Recent political activity Have added or deleted political information from their Facebook profile Have added or deleted applications that deal with politics from their Facebook profile Regarding receiving messages from political friends or fans, 44% of respondents received messages on their walls, 25.9% received direct messages, 28.9% received messages through the advertisement column, and 1.2% selected the other category (Table 4.1). Among the activities least preferred, being updated about politicians daily schedule had the lowest percentage of 14.1%, followed by politicians` criticism against each other (14.7%). As much as 28.7% of the respondents were least interested in politicians` arguments on current issues. The least preferred activity was being updated on politicians family members (42.6%). According to the findings, respondents first impressions of the politicians were 65.9% positive and 34.1% neutral. After interacting with the politicians on Facebook, 85.9% of respondents had a more positive impression of politicians while 14.1% had a less positive impression (Table 4.1). When asked what they would like to say to their politicians if they met them, 65% of respondents said they would like to interact more with politicians on Facebook, 25.5% wanted to be made deliverable promises, and 9.5% requested that issues be handled more effectively (Table 4.1). 131

152 Most participants (63.2%) claimed that they were into political parties that spread a good ideology; 27.3% preferred political parties that promised to deliver services favourable to their parties and 9.5% would go for political parties that understand their wishes and try to fulfil them. More than 66% of respondents would not mind accepting a politician s fans/friends invitation to attend the politician s campaign events, 24.8% were prone to attend the politician s family gathering while 8.7% stated that they would not be interested in meeting the politician face-to-face (Table 4.1). Table 4.1 also shows the level of social media activity where 62% of the respondents had friends on Facebook, and 38% of them had followers on Twitter. 43.4% claimed that they wanted to vote for a candidate if he or she has a good image, 27.3% would go for a candidate who promoted a better future; 19.8% would select a candidate who had fulfilled previous favours, and 9.5% would choose someone who brings good image to their political party (Table 4.1). Almost half (48.9%) of participants said their knowledge about the politician was very good; 23.4% claimed they had good knowledge about the politician, 15.4% said their knowledge was average, 10.1% said they had poor knowledge, and 2.2% claimed that they had very poor knowledge about the politician (Table 4.1). When the participants were asked whom they liked better, 80% replied thatthey preferred politicians who held cabinet posts better whereas 16.8% liked those occupying the parliamentary seats and 3.2% would not mind going for vibrant, young politicians. As for political activism, 60.6% of the respondents said that they would be voting in the coming election, 17.2% declared they would not vote in the next election, 15.8% had never 132

153 registered as voters, and 6.3% would register as voters (Table 4.1). (This is not the same liked on Facebook) Referring to Table 4.1, the largest proportion of participants in this study (71.3%), had already made up their minds who they would vote for; 41.8% were prepared to change their opinion while 29.5% said they would stick to their choice and 28.7% declared they were still undecided about who to vote for. When asked if they considered themselves active users of Facebook in politics, 82.2% of the respondents replied yes and the remainder (17.8%) said they did not consider themselves active users (Table 4.1). More than two-thirds of the respondents (71.1%) said they believed Facebook usage could determine the outcome of an election; however, 28.9% did not believe this was the case. The percentage of participants, who had included or removed any political information from their Facebook profile was70.3%, whereas those who had added or deleted applications that dealt with politics from their profiles were 29.7%. 4.3 Structural Model Analysis Hypothesised Model Hypotheses testing were able to be completed through the examination of the standardised parameter estimates and their associated t-values (Bollen, 1989). The two-tailed test of significance functioned to determine the significance of each path coefficient. Five of the direct relationships and three indirect relationship hypotheses (Moderating Testing) were harmonious with the expectations and statistically significant as has been anticipated (Hair et al., 2010; Byrne, 2001). Based on the suggestion by Anderson and Gerbing (1988), the study adopted a comprehensive, two-stage analysis, where the measurement model was first confirmed as the 133

154 confirmatory factor analysis was used, and then the structural equation modelling was performed leaning on the measurement model of exogenous variables to estimate the fit of the hypothesised model (as given in Figure 4.2), Generated Model (Figure 4.3a) and the Respecified Model (Figure 4.3b) to the data (Sentosa, Ming, Soebyakto, & Nik Kamariah, 2012; Hair et al., 2010). The hypothesised model sought to specify and test the relationships between the observed measures and their underlying constructs, giving a confirmatory assessment of the construct validity (Bentler, 1978). In the next step, the structural model examines the direct causal relationships among the latent constructs, as had been theorised (Anderson & Gerbing, 1988). As the task of building the model was accomplished through a two-step procedure, it was believed to be an improved approach to a one-step analysis (Anderson & Gerbing, 1988). Figure 4.2 Hypothesised Model between Exogenous and Endogenous Variables 134

155 4.3.2 Generated Model Based on Hair et al. (2010), the confirmatory factor analysis (CFA) of each dimension was performed to confirm the construct of the predictors of the voter s decision. The findings indicated that all exogenous constructs were related significantly to voters decision as dependent variables (Sentosa & Nik Kamariah, 2012). The findings have endorsed the model s validity, and all the hypothesised paths in the theoretical model were at the significance level of model fit (P >0.05). Composite reliability and average variance extracted (AVE) were conducted to make the confirmation of the validation of the measurements (Hair et al., 2010; Byrne, 2001). Bollen (1989) also configured the detail of CFA, including 2 nd order analysis of latent construct variables. This study conducted the 2 nd order CFA to endorse the three dimensions of party orientation; and the measurement of exogenous and endogenous variables of the structural model had pointed to the significance of the goodness of the indexes (Byrne, 2001). The Generated Model further verified the interaction of the items, variables, and the hypothesis paths, where it established significant results (Figure 4.3a). The findings of the results of the Re-specified Model demonstrate the validity of the structural model based on the hypotheses (see Figure 4.2). 135

156 Figure 4.3a Generated Model of Voters Decision 136

157 4.3.3 Re-Specified Model of Voters Decision Based on Hair et al. (2010), path analysis of the hypothesised models was tested (Figure 4.3a). In the Re-specified Model, there were five exogenous variables including three dimensions of party orientation that were hypothesised as predictors in the voters decision model. Indices evaluate the goodness-of-fit of the covariance structural models: (1) chisquare, (2) degree of freedom, (3) ratio, (4) P-value, (5) RMSEA, (6) Tucker-Lewis index (TFI), (7) Goodness-of-fit index (GFI), and (8) Comparative fit index (CFI). The most common goodness-of-fit index is the chi-square value. One major principle is that if the P- value of the chi-square statistic is greater than 0.05 (i.e., the chi-square value is not significant), then the proposed model is acceptable (Hair et al., 2009). Thus, the null hypothesis suggests that the sample covariance matrix (S) is equal to the population covariance matrix (ΣΘ) of the population of the implied model. Due to the fact that the traditional chi-square test is very sensitive to sample size, Bollen (1989) considered using the RMSEA as the principal goodness-of-fit index (GFI). A value of RMSEA of not more than 0.08 indicates a close fit and depicts a reasonable error of approximation in the population. The GFI is similar to the R -squared multiple regression coefficients because it speaks for the proportion of the observed covariance justified by the model-implied covariance (Kline 1998). The values of GFI, IFI and CFI can range between 0 and 1,with values nearing 1 to indicate a close fit between the data and the model (Jöreskog &Sörbom, 1993). With the help of a structural relations model that brought together the factor analysis, regression analysis and path analysis in this study (Bollen, 1989), the evaluation of the significance and direction of the relationships between the three exogenous and endogenous variables had been made possible. To make sure that the two purposes, the two-step procedure 137

158 suggested by Anderson and Gerbing (1988) was adopted, firstly, the measurement models were put to the test. Then, the fit of the models was appraised and then the best model selected with regard to the criteria above. Then, a second step took place to dwell on the significance of the relationships among the endogenous and exogenous variables (Figure 4.3b). The covariance matrix for the observed variables served as the input for all models (Bollen, 1989). The re-specified model also verified that all data and measurements were normally distributed (Table 4.3b). In the study, the structural equation model was a complete path model, and it was represented in a path diagram (Bollen, 1989). It was not the same as the simple path analysis in which all variables are latent variables measured by multiple indicators carrying associated error terms, additional to the residual error factor linked with the latent endogenous variables (Figure 4.3b). Structural equation modelling in the study was divided into a measurement model and a structural model (Hair et al., 2010; Byrne, 2001). The former dealt with the relationships between the measured variables and latent variables (Hypotheses 1-17). The Re-specified Model of the structural model only dealt with the relationships between latent variables (Figure 4.3b and Hypothesis 17). One of the advantages of SEM is that the latent variables can well escape the random error as the error has been estimated and eliminated and, thus, only a common variance is left (Sentosa & Nik Kamariah, 2012). 138

159 e2 e4 e5 e7 e26 e25 e24.23 e23 e22 e21 e20 PRO PRO3.95 PRO4 e31 e30 e29 e28 SAL5.95 SAL SAL3.95 SAL2 R Product MA1.95 MA MA3.95 MA5 Sales R02 R03 Market e41 LIKE1.96 e40 LIKE2.96 e38 e37 e34 e33 Party Orientation LIKE6 AP6.96 LIKE4.96 e36 AP9.94 e35 AP AP7.95 Political Activities Political Aptitude CP Candidate Profiles.58 CP4 CP Voters Decision.46 Internet Utilisation Internet Utility R VD7 VD9 VD11 CP8 e16 e18 e Standardized estimates Chi-Square : DF : 366 Ratio : P-Value : GFI :.901 TLI :.974 RMSEA :.050 INT2 e14 INT7 e10 INT9 e9 Figure 4.3b Re-Specified Model of Voters Decision 139

160 Table 4.2 Standardised Regression Weights of Re-Specified Model Endogenous Exogenous Estimate S.E. C.R. P Remark Hypothesis Objectives Sales <--- Party Orientation Significant H2 Supported Obj.6 Achieved Product <--- Party Orientation Significant H2 Supported Obj.6 Achieved Market <--- Party Orientation Significant H2 Supported Obj.6 Achieved Voters Decision <--- Political Activities Significant H1e Supported Obj.1 Achieved Voters Decision <--- Candidate Profiles Significant H1b Supported Obj.2 Achieved Voters Decision <--- Political Aptitude Significant H1d Supported Obj.3 Achieved Voters Decision <--- Internet Utilisation Significant H1a Supported Obj.4 Achieved Voters Decision <--- Party Orientation Significant H1c Supported Obj.5 Achieved Table 4.2 shows the detailed achievement of research objective 1 to research objective 6 and also acceptance of research hypothesis 1 to research hypothesis 2 as proposed in Chapter 3. The present study confirmed sales, product and market orientation as valid dimensions of party orientation. Furthermore, using path analysis, the direct positive significant influence of predictors on voters decision was also confirmed. Table 4.2 and Table 4.3 also succeeded in confirming the acceptance of hypothesis 1 and 2, as well as the achievement of objective 1 to 6 (as stated in Chapter 1). 140

161 Table 4.3 Factor Loading of Re-Specified Model Items Variable Estimate S.E. C.R. P Remark CP2 <--- Candidate Profiles Significant CP4 <--- Candidate Profiles Significant CP5 <--- Candidate Profiles Significant CP8 <--- Candidate Profiles Significant INT9 <--- Internet Utilisation Significant INT7 <--- Internet Utilisation Significant INT2 <--- Internet Utilisation Significant VD7 <--- Voters Decision Significant VD9 <--- Voters Decision Significant VD11 <--- Voters Decision Significant SAL2 <--- Sales Significant SAL3 <--- Sales Significant SAL4 <--- Sales Significant SAL5 <--- Sales Significant PRO4 <--- Product Significant PRO3 <--- Product Significant PRO2 <--- Product Significant MA5 <--- Market Significant MA3 <--- Market Significant MA2 <--- Market Significant MA1 <--- Market Significant AP6 <--- Political Aptitude Significant AP7 <--- Political Aptitude Significant AP8 <--- Political Aptitude Significant AP9 <--- Political Aptitude Significant LIKE6 <--- Political Activities Significant LIKE4 <--- Political Activities Significant LIKE2 <--- Political Activities Significant LIKE1 <--- Political Activities Significant 141

162 4.4 Competing Model using Moderating Variables Moderating Effect of Facebook Time The present study hypothesised Facebook Time as a moderating variable in the structural model. Using interaction effect analysis as suggested by Kenny and Jude (1994) and moderated structural equation modeling, this study confirmed the significant positive influence of Facebook time. Using SMARTPLS version 3.0, the competing model was configured and measured with detailed interaction as hypothesised (Table 4.4). The results show the factor loading for Internet utilisation (0.457), candidate profiles (0.586), party orientation (0.859), political aptitude (0.788) and political activities (0.655) indicating that they succeeded as predictors of Facebook time on the influence on voters decision (Figure 4.4). Figure 4.4 Moderating Effect of Facebook Time 142

163 4.4.2 Moderating Effect of Facebook Friends The present study also hypothesised Facebook Friends as a moderating variable in the structural model (Table 4.4). Using interaction effect analysis suggested by Kenny and Jude (1994) and also the use of moderated structural equation modeling, this study confirmed the positive significant influence of Facebook friends as a moderating push variable. Using SMARTPLS version 2.0, the competing model also succeeded in configuring and measuring the detailed interaction as hypothesised. The results show the factor loading for Internet utilisation (0.657), candidate profiles (0.786), party orientation (0.799), political aptitude (0.785) and political activities (0.765) indicating that they succeeded as predictors of Facebook time on the influence on voters decision (Figure 4.5). Figure 4.5 Moderating Effect of Facebook Friends 143

164 The partial least square (PLS) technique was adopted to delve into the results to assess the influence of all constructs in the framework at the same time, inclusive of the second order construct (party orientation), all exogenous variables and the moderating effect of Facebook time and friends. Tabachnick and Fidell (2007) regarded the PLS as a sustainable technique that could assess the cause and effect relationships in intricate business research. The PLS approach can assist in getting the values for latent variables for predictive reasons. PLS never tries to use a model to elaborate on the covariance of all indicators, but it lessens the variance of all dependent variables, based on the estimated parameters obtained based on the ability to reduce the dependent variables residual variance (Chin, 1998). Chin (1998a) further offered a suggestion that PLS is an estimate, indicating the latent variables with their indicators and was an outcome in a more precise analysis and their relationships. The wide adoption of the PLS is due to its capability to model latent variables in non-normal situations, with small-and-medium-sized samples. However, unlike the multiple regression PLS does not assume equal weights for all the items on a scale, because it enables each individual indicator to differ based on its contribution weight to the latent variable. According to Chin (1998a), SEM methods remain to be one of the most well-established covariance-based techniques and many researchers in social science see it as being tautologically the same as PLS. Nevertheless, Chin made a claim that PLS was an alternative method for researchers who had the interest in conducting an SEM-based analysis. PLS can be the most suitable method based on the objective of the researchers and the epistemic data view to the theory, the properties of the data in hand and also the measurement development. 144

165 Byrne (2011) pointed out that in the PLS approach, there are not that many probabilistic hypotheses, where the data is modeled by a simple or multiple regression succession, and there is no issue of identification. Furthermore, it was noted that in the SEM (LISREL) approach, estimation is performed by maximum likelihood based on the multi-normality hypotheses and permits the variance matrix to be modeled. PLS carries the potential to be one of the best techniques of analysis following its measurement scale, minimal requirements, distribution of residuals and sample size (Chin, 1998b). In contrast, Bollen (1989) contended that while PLS is prediction-oriented and would work best for accuracy prediction, supports most complex models and is quite flexible, the orientation of SEM leans more towards the estimation of parameters and is optimal for the accuracy of parameters, supports small moderately intricate models and is reliant on major assumptions. Chin (1998b) made the argument that the superiority of LISREL to PLS is explained by its ability to approximate the underlying population parameters. Nonetheless, this is less of a concern if the objective is to elaborate on the multivariance in a predictive sense. The analysis of the path is to assess the effect size ( f 2 ) to differentiate the path that contributes towards explaining the dependent variable to which they are attached (Hair et al., 2010; Byrne, 2001). Chin (1998b) stated further that the R 2 for each latent variable (LV) could be an opening point when PLS is evaluated for the structured model since the explanation of the PLS is similar to that of a traditional regression. This study used the PLS technique as the suggested theoretical framework as it had high complexity in its variables and relations, and also because the model had the intention to foresee future behaviour rather than population parameters estimation. The software utilised was Smart PLS 2.0 (Ringle, Wende, & Will, 2005). 145

166 This study recommended and managed to confirm a moderation model to test empirically and to confirm that there were positive direct relationships between Facebook Time and Facebook Friends (Hair et al., 2010; Byrne, 2001). To achieve this objective, the PLS data analysis technique was adopted (Figure 4.5). Table 4.4 also establishes the detailed results of the moderating analysis in the summary below. Table 4.4 Summary of Moderating Effect Analysis Moderating No. Exogenous Endogenous Effect P-Value Remark Achievement Achievement Variables Variable of of FB FB Hypothesis Objectives Time Friends 1 Internet Voters Significant H.3a Obj. 1 Utilisation Decision H.4a 2 Candidate Voters Significant H.3b Obj. 2 Profiles Decision H.4b 3 Party Voters Significant H.3c Obj. 3 Orientation Decision H.4c 4 Political Voters Significant H.3d Obj. 4 Aptitude Decision H.4d 5 Political Voters Significant H.3e Obj. 5 Activities Decision H.4e 4.5 Results of Hypothesis Testing In this study, the direct and indirect effects were shown through the path analysis of the variables, which is the statistical technique that examined the causal relationships between two or more variables (Sentosa & Nik Kamariah, 2012; Byrne, 2001). The technique is used primarily to comprehend the comparative strengths of direct and indirect relationships among 146

167 a set of variables ((Sentosa, Ming, Soebyakto, & Nik Kamariah, 2012; Hair et al., 2010; Byrne, 2001). It could be seen here that the path analysis is distinctive from other linear equation models. The moderated pathways (i.e., those acting as a moderating variable, Y, in the pathway X à Y à Z) could be examined in this analysis (Bollen, 1989). The multivariate procedure allows the examination of a set of relationships between two or more exogenous variables, be it continuous or discrete, and one or more endogenous variables, be it continuous or discrete (Tabachnick and Fidell, 2007). Furthermore, SEM attempts to clarify the relationships among multiple variables and checks on the structure of interrelationships expressed in several multiple regression equations (Hair et al., 2009). It is an integration of multiple regression and factor analysis. In the same way, SEM allows the investigation of sets of relationships between one or more exogenous variables (IVs), either continuous or discrete, and one or more endogenous variables (DVs), either continuous or discrete (Tabachnick & Fidell, 2007). The findings show that the latent constructs of the exogenous variables of the model remarkably had a relation to the development constructs of the Voters` Decision Model ((Sentosa, Ming, Soebyakto, & Nik Kamariah, 2012; Bryne, 2001). In specific, all hypotheses were supported, and the final structural equation model showed that the five hypothesised paths in the theoretical model were at a significant level (P>0.05). The detailed results of the hypothesis testing are as follows: 147

168 1. H1a Internet Utilisation has a significant direct influence on Voting Decision The present study verified that there was a direct positive significant influence of internet utilisation on voters decision in a future Malaysian election. Aided by the path analysis of the Re-specified Model, standardised regression weights pointed to a factor loading of 0.45, and therefore, Hypothesis 1a was supported. 2. H1b Candidate Profiles has a significant direct influence on Voting Decision The study confirmed a direct positive significant influence of candidate profiles as exogenous variables on voters decision in a future Malaysian election. Using path analysis of the Re-specified Model, standardised regression weights showed a factor loading of0.45, and Hypothesis 1b was supported. 3. H1c Party Orientation has a significant direct influence on Voting Decision The study confirmed a direct positive significant influence of party orientation on voters decision in a future Malaysian election. Using path analysis of the Re-specified Model, standardised regression weights showed a factor loading of 0.33, and Hypothesis 1c was supported. 148

169 4. H1d Political Aptitude has a significant direct influence on Voting Decision The study confirmed a direct positive significant influence of political aptitude on voters decision in a future Malaysian election. Using path analysis of the Re-specified Model, standardised regression weights showed a factor loading of0.547, and Hypothesis 1d was supported. 5. H1e Political Activities has a significant direct influence on Voting Decision The study confirmed a direct positive significant influence of political activities on voters decision in a future Malaysian election. Using path analysis of the Re-specified Model, standardised regression weights showed a factor loading of 0.457, and Hypothesis 1e was supported. 6. H2 Product, Sales and Marketing confirm the construct of Party Orientation The study confirmed a direct positive significant influence of product, sales and marketing as dimensions of party orientation. Using 2nd order CFA, it was confirmed that there would be 3 valid dimensions to measure party orientation on the voters decision in a future Malaysian election. Using path analysis of the Re-specified Model, standardised regression weights showed a factor loading of for Sales, for Product and for Market. Thus Hypothesis 2 was supported. 149

170 7. H3a Facebook Time has a moderating influence on the relationship between Internet Utilisation and Voting Decision The study confirmed a moderating effect of Facebook Time on the direct positive significant influence of internet utilisation on voters decision in a future Malaysian election. Using path analysis of moderated structural equation modeling (MSEM), the interaction effect between predictors and also standardised regression weights showed a factor loading of 0.457, and Hypothesis 3a was supported. 8. H3b Facebook Time has a moderating influence on the relationship between Candidate Profiles and Voting Decision The study confirmed a moderating effect of Facebook Time on the direct positive significant influence of candidate profiles on voters decision in a future Malaysian election. Using path analysis of moderated structural equation modeling (MSEM), the interaction effect between predictors and also standardised regression weights showed a factor loading of 0.586, and Hypothesis 3b was supported. 9. H3c Facebook Time has a moderating influence on the relationship between Party Orientation and Voting Decision The study confirmed a moderating effect of Facebook Time on the direct positive significant influence of party orientation on voters decision in a future Malaysian election. Using path analysis of moderated structural 150

171 equation modeling (MSEM), the interaction effect between predictors and also standardised regression weights showed a factor loading of 0.859,and Hypothesis 3c was supported. 10. H3d Facebook Time has a moderating influence on the relationship between Political Aptitude and Voting Decision The study confirmed a moderating effect of Facebook Time on the direct positive significant influence of political aptitude on voters decision making in a future Malaysian election. Using path analysis of moderated structural equation modeling (MSEM), the interaction effect between predictors and also standardised regression weights showed a factor loading of0.788, and Hypothesis 3d was supported. 11. H3e Facebook Time has a moderating influence on the relationship between Political Activities and Voting Decision The study confirmed a moderating effect of Facebook Time on the direct positive significant influence of political activities on voters decision in a future Malaysian election. Using path analysis of moderated structural equation modeling (MSEM), the interaction effect between predictors and also standardised regression weights showed a factor loading of 0.655, and Hypothesis 3e was supported. 151

172 12. H4a Facebook Friends has a moderating influence on the relationship between Internet Utilisation and Voting Decision The study confirmed a moderating effect of Facebook Friends on the direct positive significant influence of internet utilisation on voters decision in a future Malaysian election. Using path analysis of moderated structural equation modeling (MSEM), the interaction effect between predictors and also standardised regression weights showed a factor loading of 0.657, and Hypothesis 4a was supported. 13. H4b Facebook Friends has a moderating influence on the relationship between Candidate Profiles and Voting Decision The study confirmed a moderating effect of Facebook Friends on the direct positive significant influence of candidate profiles on voters decision in a future Malaysian election. Using path analysis of moderated structural equation modeling (MSEM), the interaction effect between predictors and also standardised regression weights showed a factor loading of 0.786, and Hypothesis 4b was supported. 14. H4c Facebook Friends has a moderating influence on the relationship between Party Orientation and Voting Decision The study confirmed a moderating effect of Facebook Friends on the direct positive significant influence of party orientation on voters decision in a future Malaysian election. Using path analysis of moderated 152

173 structural equation modeling (MSEM), the interaction effect between predictors and also standardised regression weights showed a factor loading of 0.799, and Hypothesis 4c was supported. 15. H4d Facebook Friends has a moderating influence on the relationship between Political Aptitude and Voting Decision The study confirmed a moderating effect of Facebook Friends on the direct positive significant influence of political aptitude on voters decision in a future Malaysian election. Using path analysis of moderated structural equation modeling (MSEM), the interaction effect between predictors and also standardised regression weights showed a factor loading of 0.785, and Hypothesis 4d was supported. 16. H4e Facebook Friends has a moderating influence on the relationship between Political Activities and Voting Decision The study confirmed a moderating effect of Facebook Friends on the direct positive significant influence of political activities on voters decision in a future Malaysian election. Using path analysis of moderated structural equation modeling (MSEM), the interaction effect between predictors and also standardised regression weights showed a factor loading of 0.765, and Hypothesis 4e was supported. 153

174 17. H5 Internet Utilisation, Candidate Profiles, Party Orientation, Political Aptitude, Political Activities, moderated by Facebook Time and Facebook Friends confirm significant interaction in the Structural Model. Table 4.3 and Table 4.4 highlight the full account of the goodness-ofmodel-fit of the structural model (Hair et al., 2010; Byrne, 2001). The testing of the endogenous structural models and the testing of endogenous variable shed light on the significance of P level (P > 0.05), GFI (GFI > 0.90) which fulfilled the RMSEA criteria (not more than 0.08). A comparison between the generated model (Figure 4.3a) and the Respecified Model (Figure 4.3b) confirmed that the final hypothesised model (i.e., H5) matched the Re-specified Model, where the probability (P=0.055), RMSEA (0.048) and GFI (0.901) achieved the significant level of the goodness-of-fit index of the model (Sentosa et al., 2012). Thus, Hypothesis 5 was supported enabling the configuration of a valid and significant interaction of Internet Utilisation, Candidate Profiles, Party Orientation, Political Aptitude, and Political Activities on voters decision (Figure 4.5). The summary of the hypothesis testing is given in Table 4.5. The present study also confirmed that all indexes on the journey to the structural model achieved the fit criteria (P-value > 0.05; GFI >0.9, TLI >0.9 and RMSEA < 0.08), and achievement of square multiple correlation (SMC) of the Re-Specified model. 154

175 Table 4.5 Summary of Hypothesis Testing No. Hypothesis Statement Factor Loading Significant Test Results Hy1a Hy1b Hy1c Hy1d Hy1e Hy2 Hy3a Hy3b Hy3c Hy3d Hy3e Hy4a Hy4b Hy4c Hy4d Hy4e H5 Internet Utilisation has a direct significant influence on Voting Decision Candidate Profiles has a direct significant influence on Voting Decision Party Orientation has a direct significant influence on Voting Decision Political Aptitude has a direct significant influence on Voting Decision Political Activities has a direct significant influence on Voting Decision Product, Sales and Marketing party orientation has a direct significant influence on Voting Decision Sales oriented party has a direct significant influence on Voting Decision Product oriented party has a direct significant influence on Voting Decision Marketing oriented party has a direct significant influence on Voting Decision Facebook Time has a moderating influence on the relationship between Internet Utilisation and Voting Decision Facebook Time has a moderating influence on the relationship between Candidate Profiles and Voting Decision Facebook Time has a moderating influence on the relationship between Party Orientation and Voting Decision Facebook Time has a moderating influence on the relationship between Political Aptitude and Voting Decision Facebook Time has a moderating influence on the relationship between Political Activities and Voting Decision Facebook Friends has a moderating influence on the relationship between Internet Utilisation and Voting Decision Facebook Friends has a moderating influence on the relationship between Candidate Profiles and Voting Decision Facebook Friends has a moderating influence on the relationship between Party Orientation and Voting Decision Facebook Friends has a moderating influence on the relationship between Political Aptitude and Voting Decision Facebook Friends has a moderating influence on the relationship between Political Activities and Voting Decision Internet Utilisation, Candidate Profiles, Party Orientation, Political Aptitude, Political Activities, moderated by Facebook Time and Facebook Friends confirm significant interaction in the Structural Model Supported Supported Supported Supported Supported Supported Supported Supported Supported Supported Supported Supported Supported Supported Supported Supported Supported Supported Supported 155

176 4.6 Achievement of Research Objectives Having evaluated the measurement scales for the reliability and validity of research measurement, the construct assessment of the study was carried out using the confirmatory factor analysis and a measurement model (Byrne, 2001). Scale purification was undertaken using the CFA of individual latent constructs to find out more about the theoretical relationships between the observable (measurable) indicators and the theoretical constructs and to ascertain the extent to which the data set fitted a hypothesised CFA model (Sentosa & Nik Kamariah, 2012). CFA was also conducted to examine the properties of measures in the aspects of testing convergent and discriminant validity and the reliability properties of the measures to identify internal consistency and adequate fit of scale items (Kline, 1998). To study the fit of the observed variables to the latent variables, the evaluation of the overall model fit indices was performed. The measurement models of both exogenous and endogenous constructs were also constructed to evaluate the unidimensionality of the measurement scale. Items with low factor loadings and high residuals were removed via an iterative process to have a better model fit and more reliable constructs to further analyse the SEM. After the measures were estimated, the values of covariance among exogenous constructs were examined to identify any multicollinearity issues that might occur among predictors in the model (Byrne, 2001). From the results established in the preceding section, evidently, all the research objectives of the study were achieved. Research Objective 1 In terms of whether Facebook fans as voters in Malaysia anticipate future return/rewards for their votes judging from the last general election in Malaysia, a lot of young computer-savvy voters tended to have voted the opposition party simply to teach the ruling party a good lesson and to give them a strict warning to expect the 156

177 worst if they do not change. Thus, they are not concerned with future consideration for their votes. Or perhaps could the prospect of change in the future can be accounted for? This can be a modified version of PMET. Using the re-specified model, objective 1 was fulfilled. Research Objective 2 To determine whether Facebook fans as voters in Malaysia have voted for candidates that have done them a favour in the past. In one by-election in Port Dickson, another ex-chief minister was elected by the voters (although he had retired 8 years earlier) simply because he had done so much for the voters in the past. The main difference between this case and the earlier one was that this ex-chief minister was contesting under the ruling party. Using re-specified model, the objective 2 was achieved. Research Objective 3 To determine whether Facebook fans as voters in Malaysia regard the candidate (background and standing) as more important than the future promises or favours they made in the past? Could the voters take into account the capability of the candidate to deliver his promises in the future when casting their votes? In another case, a young ex-teacher contested under the ruling party s banner against a heavyweight corporate and social champion. He won the by-election based on the reliance on the existing ruling party to deliver the promises. However, during the last general election, he lost against the same opponent, which subsequently resulted in the ruling party being toppled in the state. The opponent subsequently became the chief minister of the state. What happened here? Were the voters able to detect the trend rationally and switch their support? Or they took a risk trying to change the trend and the government? Using re-specified model, the objective 3 was achieved. 157

178 Research Objective 4 To determine whether Facebook fans as voters in Malaysia regard the tangible value of promises as more influential than intangible ones. The promises and ability to honour them must be taken into account in their voting consideration. Could the above case be inferred that PMET actually worked here in that the voters cast their votes based on the assurance of the person who had the power to deliver the consideration in the future? Using the re-specified model, objective 4 was achieved. Research Objective 5 The objective sought to ascertain whether or not Facebook fans as voters in Malaysia were greatly influenced by information on the Internet in their voting behaviour in the Malaysian election (2008). The fact that the more they were exposed to the campaign blog and Facebook, the more likely they would vote for the opposition party indicates that they were just expressing their anger through their voting behaviour. Could they lack the rationality in their voting behaviour and hence dismiss the PMET consideration? The next general election is expected to be very important because the voters are now smart enough to accept a broader perspective and consideration in their voting behaviour. The young voters are crucial, especially as they constitute 40% of the voters. The growing importance of the Internet as a political marketing tool will also divert the attention towards more Internet-savvy younger generation. Using the re-specified model, the present study confirmed the achievement of objective

179 Research Objective 6 To determine whether Facebook fans as voters in Malaysia regard party image as more crucial than the candidates background. In a case, a young candidate, whose experience was just in taking amateur video of a private conversation between a lawyer and the so-called prime minister of the day had been elected despite the fact that the opponent was a goliath in politics representing the government of the day. What are the main considerations taken in casting the votes? Again, have the voters been pro-actively working on changing the trend and hence looking at other considerations than suggested by PMET in their voting behaviour? Using re-specified model, the objective 6 was achieved. In summary, throughout the analysis of all the research variables, all the research objectives were achieved, and all the research questions were answered. This was done by analysing the relationship between variables in the model. Table 4.3, Table 4.4 and Table 4.5 showed how each question in a variable had been taken into account in determining the relationships between variables. SEM uses a unique method with respect to consolidating the factors from each variable in the analysis. 4.7 Model Validation using Current Election Data (GE13) Confirmatory Factor Analysis (CFA) was conducted to configure and confirm the variables and detailed measurements as hypothesised (Sentosa et al., 2012; Hair et al., 2010). Using data from the latest election (GE13), five 1 st order CFA variables (Internet Utilisation, Candidate Profiles, Political Aptitude and Voters Like) and one 2 nd order CFA (Party Orientation) with three dimensions: Party Orientation, Marketing Orientation and Sales Orientation were investigated. 1 st order CFA also validates Voters Decision as an endogenous variable. The Goodness of Fit indexes (Chi-Square, Degree of Freedom, Ratio, P -Value, GFI, 159

180 TLI and RMSEA) were employed as indicators to confirm the significant model with valid standardised regression weights (factor loading > 0.5) and low level of standard error (< 10%). 4.8 CFA of Exogenous Variables A. CFA of Internet Utilisation Internet Utilisation as an exogenous variable was measured and validated using confirmatory factor analysis (CFA). 13 items were validated using 505 respondents as the sample (Figure 4.6a). The standardised regression weights show the convergent reliability of 10 valid items (factor loading more than 0.5). The CFA also confirmed the significant indexes of the CFA model (Figure 4.6b). Figure 4.6b and Table 4.6 confirmed the validation of Internet Utilisation as one of the political marketing factors. Using data from a sample of 150 respondents who voted in the PRU13, the present study succeeded in validating 9 items of Internet utilisation (Table 4.6). Table 4.6 Standardised Regression Weights of Internet Utilisation Items Variable Estimate S.E. C.R. P Remark Utilisation2 <--- Internet Utilisation Significant Utilisation3 <--- Internet Utilisation Significant Utilisation4 <--- Internet Utilisation Significant Utilisation5 <--- Internet Utilisation Significant Utilisation6 <--- Internet Utilisation Significant Utilisation7 <--- Internet Utilisation Significant Utilisation8 <--- Internet Utilisation Significant Utilisation9 <--- Internet Utilisation Significant Utilisation10 <--- Internet Utilisation Significant 160

181 Figure 4.6a CFA of Internet Utilisation 161

182 Figure 4.6b CFA of Internet Utilisation 162

183 B. CFA of Candidate Profiles The CFA of candidate profiles was conducted by means of constraining the original 9 items of the endogenous constructs based on the 9 questions related to the candidate profiles. Taking the modification index and error elimination as the basis, the CFA of candidate profiles confirmed that there were 9 valid items left in the model development, (Figure 4.7). The CFA of candidate profiles constructs yielded a relatively good fit as shown by the goodness-of-fit indices such as CMIN/df ratio (<2); P-value (>0.05); the Goodness-of-Fit Index (GFI) of >.90; and RMSEA values of less than.08 (<. 08), (Table 4.7). Table 4.7 Standardised Regression Weights of Candidate Profiles Items Variable Estimate S.E. C.R. P Remark Profile1 <--- Candidate Profiles Significant Profile2 <--- Candidate Profiles Significant Profile3 <--- Candidate Profiles Significant Profile4 <--- Candidate Profiles Significant Profile5 <--- Candidate Profiles Significant Profile6 <--- Candidate Profiles Significant Profile7 <--- Candidate Profiles Significant Profile8 <--- Candidate Profiles Significant Profile9 <--- Candidate Profiles Significant 163

184 Figure 4.7 CFA of Candidate Profiles 164

185 C. CFA of Party Orientation The 2nd order CFA of the party orientation (Figure 4.8a) for GE13 was conducted through the act of constraining the original 15 items of the endogenous constructs (Hair et al., 2010; Byrne, 2001). Leaning on the modification index and error removal, the CFA of party orientation confirmed that there were 13 valid items in the development of the model (Figure 4.8b). The 2nd order CFA of the party orientation construct generated a relatively good fit as indicated by the goodness-of-fit indices such as CMIN/df ratio (<2); P-value (>0.05); the Goodness-of-Fit Index (GFI) of >.90; and RMSEA values of not more than.08 (<. 08), (Table 4.8). Figure 4.8a CFA of Party Orientation 165

186 Figure 4.8b CFA of Party Orientation for GE13 166

187 Table 4.8 Standardised Regression Weights of Party Orientation (2ndOrder CFA) Items Variable Estimate S.E. C.R. P Label Sales <--- Party Orientation Significant Marketing <--- Party Orientation Significant Product <--- Party Orientation Significant Product2 <--- Product Significant Product3 <--- Product Significant Product4 <--- Product Significant Product5 <--- Product Significant Sales2 <--- Sales Significant Sales3 <--- Sales Significant Sales4 <--- Sales Significant Sales5 <--- Sales Significant Marketing5 <--- Marketing Significant Marketing4 <--- Marketing Significant Marketing3 <--- Marketing Significant Marketing2 <--- Marketing Significant Marketing1 <--- Marketing Significant 167

188 D. CFA of Political Aptitude The CFA of political aptitude was conducted by constraining the original 19 items of the endogenous constructs (Figure 4.9a). With the modification index and error elimination (Bollen, 1989) serving as the basis, the CFA of political aptitude confirmed that there were 8 valid items in the model development (Figure 4.9b and Table 4.9). Table 4.9 Standardised Regression Weights of Political Aptitude Items Variable Estimate S.E. C.R. P Label Aptitude5 <--- Political Aptitude Significant Aptitude6 <--- Political Aptitude Significant Aptitude7 <--- Political Aptitude Significant Aptitude8 <--- Political Aptitude Significant Aptitude9 <--- Political Aptitude Significant Aptitude10 <--- Political Aptitude Significant Aptitude11 <--- Political Aptitude Significant Aptitude12 <--- Political Aptitude Significant 168

189 Figure 4.9a CFA of Political Aptitude 169

190 Figure 4.9b CFA of Political Aptitude 170

191 E. CFA of Political Activities The CFA of political activities was conducted by constraining the original 10 items of the endogenous constructs (Figure 4.10a). From the modification index and error elimination (Bollen, 1989), the CFA of political activities confirmed that there were 8 valid items in the model development (Figure 4.10b and Table 4.10). The CFA of the candidate profiles constructs yielded a relatively good fit as demonstrated by the goodness-of-fit indices such as CMIN/df ratio (<2); P-value (>0.05); the Goodness-of-Fit Index (GFI) of >.90; and RMSEA values of less than.08 (<. 08). Table 4.10 Standardised Regression Weights of Voters Like Items Variable Estimate S.E. C.R. P Remark Like1 <--- Voters Like Significant Like2 <--- Voters Like Significant Like3 <--- Voters Like Significant Like4 <--- Voters Like Significant Like5 <--- Voters Like Significant Like6 <--- Voters Like Significant Like8 <--- Voters Like Significant Like9 <--- Voters Like Significant 171

192 Figure 4.10a CFA of Political Activities (Voters Like) 172

193 Figure 4.10b CFA of Political Activities (Voters Like) 173

194 4.9 CFA of Endogenous Variable (Voters Decision) The CFA of voters decision was carried out by constraining the original 11 items of the endogenous constructs (Figure 4.11a). Depending on the modification index and error elimination (Bollen, 1989), the CFA of voters decision verified that there were 9 valid items in the development of the model (Figure 4.11b). The CFA of the candidate profiles construct yielded a relatively good fit indicated by the goodness-of-fit indices such as CMIN/DF ratio (<2); P-value (>0.05); the Goodness-of-Fit Index (GFI) of >.90; and RMSEA values of less than.08 (<. 08), (Hair et al., 2010; Byrne, 2001) (Table 4.11). Table 4.11 Standardise Regressions Weight of Voters Decision Items Variable Estimate S.E. C.R. P Label Decision3 <--- Voters Decision Significant Decision4 <--- Voters Decision Significant Decision5 <--- Voters Decision Significant Decision6 <--- Voters Decision Significant Decision7 <--- Voters Decision Significant Decision8 <--- Voters Decision Significant Decision9 <--- Voters Decision Significant 174

195 Figure 4.11a CFA of Voters Decision 175

196 Figure 4.11b CFA of Voters Decision 176

197 4.10 Summary In the present study, the Root-Mean Squared Error of Approximation (RMSEA) was used to examine and confirm the hypothesised model (Hair et al., 2010; Byrne, 2001; Bollen, 1989). RMSEA indicated the extent to which the residuals in the model differ from zero. In general, models with RMSEA of less than 0.05 are considered to reflect a good fit and models with RMSEA of between.05 and.08 are considered to have a fair fit to the data (Browne & Cudeck, 1993). The significant level of factor loading interactions between all variables in the model confirmed the moderating effects of Facebook Time and Facebook Friends on the relationship between political marketing factors (internet utilisation, candidate profiles, party orientation, political aptitude and political activities). Path analysis and hypothesis testing showed the total of direct and indirect effects of variables interactions and indicated the goodness of model fit (GOF) of the re-specified model (Sentosa et al., 2012; Hair et al., 2010). The results of SEM in assessing the validity or the empirical relationship between constructs of political marketing were positively related to voters decision, as hypothesised. The variables of internet utilisation, candidate profiles, party orientation, political aptitude and political activities were found to be positively related to voters decision for two sets of data for the 2008 election and validated with data from the 2013 election. Through interaction effect analysis, Facebook Time and Facebook Friends were confirmed as moderators on the relationship between political marketing factors and voters decision. Therefore, the objectives of the present study had been achieved. The next chapters will further discuss the findings, implications of the findings to practice, limitations of the study, and suggestions for future research. 177

198 CHAPTER 5 VALIDATION AND RELIABILITY TESTS 5.1 Data Screening Multivariate Outliers There is one suggestion by J. Cohen (1983) and P. Cohen (1983) that, as a matter of principle, the multivariate outliers must be equivalent to the Mahalanobis Distance and it must be higher than chi-square value. Hair et al. (2009) proposed that Mahalanobis distance should not surpass the critical chi-squared value with degrees of freedom equal to the number of predictors and alpha = 0.001, and Tabachnick and Fidell (2007) suggested that the Mahalanobis distance should be defined as χ 2 statistic with the degree of freedom equal to the number of items. One recommendation is that a criterion of P < serves to evaluate whether or not a case is judged to be a multivariate outlier. In this study, a test for multivariate outliers was carried out using the techniques that had been detailed by Tabachnick and Fidell (2007) based on the compilation of data of 550 respondents. The Mahalanobis distance was calculated based on 103 items. The criterion of P < and critical value of χ 2 = were adopted. The tests conducted identified 37 cases with Mahalanobis values above The Mahalanobis distance was able to identify the multivariate outliers. These 37 cases identified as outliers were then deleted. 8 more questionnaires were unusable because there were found to be incomplete. This study verified that 505 cases (respondents) were used to be analysed further Normality Test of Measurements Normality of variables was assessed by these two methods- skewness and kurtosis. On average, skewness and kurtosis values were smaller than one, and the deviations from 178

199 normality were minimal. Logarithm, square, and square root transformations of the data did not reduce these deviations, and with this in mind, the data were retained in their original form. Tabachnick and Fidell (2007) stated that the impact of small skewness and kurtosis deviations from zero disappears in a sample of 200 or more. As the sample size in this study was 505, slight skewness and kurtosis were deemed not to determine the overall analyses and results. Linearity was assessed with regard to the residual plots from the regression analyses. Normally and independently distributed residuals showed the independence of error terms. Bivariate scatterplots between variables showed that the relationships were homoscedastic and evenly distributed (Hair et al., 2009). When combined, these analyses indicated that the data were reliable and appropriate to be studied further. The normality of the distribution of variable scores were investigated. The value of skewness and kurtosis were calculated to seek identification of the distribution of scores for each item in the variables. The value of skewness can also be obtained when the standard error of the skewness is divided (Tabachnick & Fidell, 2007). A total of 17 items were found to be non-normal. Table 5.1a and Table 5.1b below show the identification of the non-normal items among the items. 179

200 Table 5.1a Normality Test and Identification of Non-Normal of Measurements Value of Original Items Normal Normal Items Items Items Std. Error Std. Error Skewness (Skewness / Kurtosis (Kurtosis / of Skewness of Kurtosis Std. Error Std. Error < 2.58) < 7) AP6 (sqrt) AP7 (sqrt) CP4 (sqrt) PRO5 (sqrt) SAL4 (sqrt) MA1 (sqrt) MA2 (sqrt) MA3 (sqrt) LIKE1 (sqrt) LIKE4 (sqrt) LIKE5 (sqrt) INT4 (ln10) INT5 (sqrt) VD7 (sqrt) VD8 (sqrt) VD9 (sqrt) VD10 (sqrt) Note: Sqrt = Square Root; Ln10 = Log

201 Table 5.1b. Final Transformation of Non-Normal Items Value of Normal Items Normal Items Normal Items Items Std. Error of Std. Error of Skewness (Skewness / Std. Kurtosis (Kurtosis / Std. Skewness Kurtosis Error < 2.58) Error < 7) AP6 (sqrt) AP7 (sqrt) CP4 (sqrt) PRO5 (sqrt) SAL4 (sqrt) MA1 (sqrt) MA2 (sqrt) MA3 (sqrt) LIKE1 (sqrt) LIKE4 (sqrt) LIKE5 (sqrt) INT4 (ln10) INT5 (sqrt) VD7 (sqrt) VD8 (sqrt) VD9 (sqrt) VD10 (sqrt) Note: Sqrt = Square Root; Ln10 = Log 10. First of all, the scores for these variables were reflected, and then a square root transformation was applied. To retain the rank order of the original raw scores, the scores were reflected again to produce normal data of a new variable. This yielded a z-score or a critical ratio of Thus, all constructs in the variables were normalised as the critical score of 2.58 fell within the skewness value lower than 2.0, and kurtosis value smaller than 7.0 as suggested by (Tabachnick and Fidell, 2007). Conclusively, all the given variables were distributed normally, which is as given in Table 5.1b. 181

202 5.2 Pilot Test Results (N=50) This section covers the aspects of the reliability of the data and the variables. The reliability of valid measures depends on the internal consistency and the dimensionality of the measurement scales (Bollen, 1989). With the establishment of the reliability and validity of measures, the researcher makes the errors minimal. The reliability of the measures was evaluated by running a test on the internal consistency and the reliability of the measurement scale prior to the item analysis (Hair, Black, Babin, Anderson, & Tatham, 2010). A pilot test of the questionnaire allows the researcher to discover how well the survey instrument works. As expected, the pre-test identified several problems and some confusion on the questionnaire content and understanding of items and sentences. The results of these pretests led to a revision of the questionnaire format, words and meanings, reducing the words in the questions. SPSS version was employed to measure the reliability of the constructs. Cronbach s alpha estimated the measures reliability. Cronbach s Alpha coefficient gauges on the scale, the correlation of scale items and the strength of relationships between one item and others. There are various schools of thought on Cronbach s Alpha coefficient output. While Hair et al. (2009) deemed alpha values > 0.6 acceptable, Nunnally, (1978) regarded > 0.7 to be adequate, and Tabachnick and Fidel (2007) regarded 0.80 as being acceptable, while 0.90 is thought of as excellent. The pilot test section showed the alpha values for the constructs that had been established in this study. 182

203 5.2.1 Reliability of Internet Utilisation This section taps into the reliability of the data and the variables. The reliability of valid measures relies on the internal consistency and the dimensionality of measurement scales (Bollen, 1989). The reliability of the measures was assessed by examining the internal consistency and the reliability of the measurement scale before assigning the items to the validation analysis (Table 5.2a). There were 13 items that were connected to the questions on Internet utilisation being tested as the measurements of the Internet utilisation variable. Table 5.2a Reliability Statistics of Internet Utilisation Variable Cronbach s Alpha No. of Items Remark Internet High Reliability Utilisation Based on Table 5.2a, the present study confirmed a high level of reliability of Internet Utilisation with a number of item measurements is 13. This study also configured a detail reliability test for each item. It confirmed a high level of reliability for items number 1 until item number 13 (Table 5.2b). The consistency of indicators had a significant interaction as a reliable measurement of Internet Utilisation. The reliability test was useful to confirm all measurements of variable and all indicators were involved further with confirmatory factor analysis (CFA) process until hypothesis testing 183

204 Table 5.2b Item-Total Statistics of Internet Utilisation Items Cronbach s Alpha Corrected Item- Remark Total Correlation Utilisation High Reliability Utilisation High Reliability Utilisation High Reliability Utilisation High Reliability Utilisation High Reliability Utilisation High Reliability Utilisation High Reliability Utilisation High Reliability Utilisation High Reliability Utilisation High Reliability Utilisation High Reliability Utilisation High Reliability Utilisation High Reliability Reliability of Candidate Profiles The pilot test sought to examine the properties of measurement scales and the items to obtain the overall index of the internal consistency of the scales by taking away the problematic items (Hair et al., 2009). Cronbach's alpha reliability test functioned to test the internal consistency of the measurement. Referring to Sekaran and Bougie (2010), the value of alpha between ±. 92 and ±. 95 represented the moderate correlation, while alpha greater than ±. 71 indicates that there was a high correlation (Sekaran & Bougie, 2010). Table 5.3a and Table 5.3b illustrate the detailed reliability test of candidate profiles with 9 reliable items related to questions concerning the candidates profiles as measurements. The test had to be repeated on each variable used in the model. 184

205 Table 5.3a Reliability Statistics of Candidate Profiles Variable Cronbach s Alpha No. of Items Remark Candidate Profile High Reliability Based on Table 5.3a, the present study confirmed a high level of reliability of Candidate Profile with a number of item measurements is 9. This study also configured a detail reliability test for each item. It confirmed a high level of reliability for items number 1 until item number 9 (Table 5.3b). The consistency of indicators had a significant interaction as a reliable measurement of Candidate Profile. The reliability test was useful to confirm all measurements of variable and all indicators were involved further confirmatory factor analysis (CFA) process until hypothesis testing. Table 5.3b Item-Total Statistics of Candidate Profiles Items Cronbach s Alpha Corrected Item- Remark Total Correlation Profile High Reliability Profile High Reliability Profile High Reliability Profile High Reliability Profile High Reliability Profile High Reliability Profile High Reliability Profile High Reliability Profile High Reliability Babbie (1998) pointed out that reliability concerns with whether a particular technique applied to the same object in repetition, would generate the same result each time. It is the extent to which measures are free from, error and therefore they produce consistent results (Zikmund, 2003). The reliability of a measure can be confirmed when both consistency and reliability are able to be tested. 185

206 Reliability is a type of association correlating a variable with itself, usually when evaluating inter-rater similarity on a variable. It is the correlation of an item, scale, or instrument with a hypothetical one that genuinely measures what it is supposed to measure. On the other hand, validity is referred as the ability of a scale or measuring instrument to measure what it is intended to measure (Zikmund, 2003). The reliability of the pilot test data (N=50) concerns with the ability of a measure to be consistent, while validity refers to the extent to which scores can precisely define a construct. Content validity emphasises the content of the test items and their relation to the intended test domain; construct validity sheds more light on the abstract construct that the test actually intends to measure, with the test item content or other factors deemed irrelevant (Sekaran &Bougie, 2010), while related validity refers to the degree of correlation between the measure and the criterion Reliability of Party Orientation As hypothesised, there were three dimensions (Sales, Product and Market) of the construct of party orientation. Sales orientation was measured with 5 items, and the same number of items also measured Product and Market orientation. As illustrated in Table 5.4a to Table 5.7b, Cronbach s Alpha confirmed the moderate and high levels of the reliability test of measurements. Table 5.4a Reliability Statistics of Party Orientation Variable Cronbach s Alpha No. of Items Remark Party Orientation High Reliability Based on Table 5.4a, the present study confirmed a high level of reliability of Party Orientation with a number of item measurements were 7 i.e. the related to questions on party orientation. This study also configured a detail reliability test for each item. It confirmed a 186

207 high level of reliability for items number 1 until item number 7 (Table 5.4b). The consistency of indicators had a significant interaction as a reliable measurement of Party Orientation. The reliability test was useful to confirm all measurements of variable and all indicators were involved further with confirmatory factor analysis (CFA) process until hypothesis testing Table 5.4b Item-Total Statistics of Party Orientation Items Cronbach s Alpha Corrected Item- Remark Total Correlation Orientation Moderate Reliability Orientation High Reliability Orientation High Reliability Orientation Moderate Reliability Orientation Moderate Reliability Orientation High Reliability Orientation High Reliability A. Product Orientation Based on Table 5.5a, the present study confirmed a high level of reliability of Product Orientation with a number of item measurements is, i.e. related to questions on product orientation. This study also configured a detail reliability test for each item. It confirmed a high level of reliability for items number 1 until item number 5 (Table 5.5b). The consistency of indicators had a significant interaction with reliable measurements of Product Orientation. The reliability test was useful to confirm all measurements of variable and all indicators were involved further with confirmatory factor analysis (CFA) of product orientation process until hypothesis testing. 187

208 Table 5.5a Reliability Statistics of Product Orientation Variable Cronbach s Alpha No. of Items Remark Product Orientation High Reliability Table 5.5b Item-Total Statistics of Product Orientation Items Cronbach s Alpha Corrected Item- Remark Total Correlation Product High Reliability Product High Reliability Product Moderate Reliability Product Moderate Reliability Product Moderate Reliability B. Sales Orientation Based on Table 5.6a, the present study confirmed a high level of reliability of Sales Orientation with a number of item measurements is 5, i.e. related to questions on Sales orientation. This study also configured a detail reliability test for each item. It confirmed a high level of reliability for items number 1 until item number 5 (Table 5.6b). The consistency of indicators had a significant interaction with reliable measurements of Sales Orientation. The reliability test was useful to confirm all measurements of variable and all indicators were involved further with confirmatory factor analysis (CFA) process until hypothesis testing of the measurement model and structural model (Table 5.6b). Table 5.6a Reliability Statistics of Sales Orientation Variable Cronbach s Alpha No. of Items Remark Sales Orientation High Reliability 188

209 Table 5.6b Item-Total Statistics of Sales Orientation Items Cronbach s Alpha Corrected Item- Remark Total Correlation Sales Moderate Reliability Sales High Reliability Sales Moderate Reliability Sales High Reliability Sales High Reliability C. Marketing Orientation Based on Table 5.7a, the present study confirmed a high level of reliability of Marketing Orientation with a number of item measurements is 5, i.e. related to questions of marketing orientation. This study also configured a detail reliability test for each item. It confirmed a high level of reliability for items number 1 until item number 5 (Table 5.7b). The consistency of indicators had a significant interaction as a reliable measurement of Marketing Orientation. The reliability test was useful to confirm all measurements of variable and all indicators were involved further with confirmatory factor analysis (CFA) process until hypothesis testing of variables and measurements. Table 5.7a Reliability Statistics of Marketing Orientation Variable Cronbach s Alpha No. of Items Remark Marketing Orientation High Reliability Table 5.7b Item-Total Statistics of Marketing Orientation Items Cronbach s Alpha Corrected Item- Remark Total Correlation Marketing Moderate Reliability Marketing High Reliability Marketing Moderate Reliability Marketing High Reliability Marketing High Reliability 189

210 5.2.4 Reliability of Political Aptitude Table 5.8b confirms that a moderate level ( ) of reliability was obtained for political aptitude. There were 19 items hypothesized as measurement of political aptitude i.e. related to questions on political aptitude. Table 5.8a Reliability Statistics of Political Aptitude Variable Cronbach s Alpha No. of Items Remark Political Aptitude Moderate Reliability Table 5.8b Item-Total Statistics of Political Aptitude Items Cronbach s Alpha Corrected Item-Total Remark Correlation Aptitude Moderate Reliability Aptitude Moderate Reliability Aptitude Moderate Reliability Aptitude Moderate Reliability Aptitude Moderate Reliability Aptitude Moderate Reliability Aptitude Moderate Reliability Aptitude Moderate Reliability Aptitude Moderate Reliability Aptitude Moderate Reliability Aptitude Moderate Reliability Aptitude Moderate Reliability Aptitude Moderate Reliability Aptitude Moderate Reliability Aptitude Moderate Reliability Aptitude Moderate Reliability Aptitude Moderate Reliability Aptitude Moderate Reliability Aptitude Moderate Reliability Based on Table 5.8a, the present study confirmed a moderate level of reliability of Political Aptitude with a number of item measurements were 19. This study also configured a detail 190

211 reliability test for each item. It confirmed a high level of reliability for items number 1 until item number 19 (Table 5.8b). The consistency of indicators had a significant interaction as a reliable measurement of Political Aptitude. The reliability test was useful to confirm all measurements of variable and all indicators were involved further with confirmatory factor analysis (CFA) process until hypothesis testing of the structural model Reliability of Political Activity (Voters Like) Cronbach's alpha reliability test examined the internal consistency of the measurement. The value of alpha between ±. 41 and ±. 70 means a moderate correlation while alpha greater than ±.71 means a high correlation (Sekaran &Bougie, 2010). Political activities were measured with a 10-item construct. The results confirmed that the Cronbach s alpha was at moderate and high levels. Table 5.9a Reliability Statistics of Political Activity (Voters Like) Variable Cronbach s Alpha No. of Items Remark Political Activity (Voters Like) High Reliability Based on Table 5.9a, the present study confirmed a high level of reliability of Political Activity (Voters Like) with a number of item measurements were 10. This study also configured a detail reliability test for each item. It confirmed a high level of reliability for items number 1 until item number 10 (Table 5.9b). The consistency of indicators had a significant interaction with reliable measurements of Political Activity (Voters Like). The reliability test was useful to confirm all measurements of variable and all indicators were involved further with confirmatory factor analysis (CFA) process until hypothesis testing. 191

212 Table 5.9b Item-Total Statistics of Political Activity (Voters Like) Items Cronbach s Alpha Corrected Item- Remark Total Correlation Like High Reliability Like Moderate Reliability Like Moderate Reliability Like Moderate Reliability Like Moderate Reliability Like High Reliability Like Moderate Reliability Like Moderate Reliability Like Moderate Reliability Like High Reliability Reliability of Voters Decision There were 11 items measured with 11 questions, i.e. related to questions on voters decision. Table 5.10a and Table 5.10b confirm the high level of reliability test as Cronbach s alpha level was more than 0.8. Table 5.10a Reliability Statistics of Voters Decision Variable Cronbach s Alpha No. of Items Remark Voters Decision High Reliability Based on Table 5.10a, the present study confirmed a high level of reliability of Voters Decision with a number of item measurements were 11. This study also configured a detail reliability test for each item. It confirmed a high level of reliability for items number 1 until item number 11 (Table 5.10b). The consistency of indicators had a significant interaction as a reliable measurement of Voters Decision. The reliability test was useful to confirm all measurements of variable and all indicators were involved further with confirmatory factor analysis (CFA) process until hypothesis testing of the structural and measurement model. 192

213 Table 5.10b Item-Total Statistics of Voters Decision Items Cronbach s Alpha Corrected Item- Remark Total Correlation Decision High Reliability Decision High Reliability Decision High Reliability Decision High Reliability Decision High Reliability Decision High Reliability Decision High Reliability Decision High Reliability Decision High Reliability Decision High Reliability Decision High Reliability The pilot test, using 50 respondents, was conducted to measure the internal consistency of and accuracy of measurements. It was shown that the Cronbach s alpha loading of all constructs had values more than 0.7. It could be concluded that all constructs and measurements were valid. This was indicated by the standardised reliability, which is significant (Hair et al., 2009). Furthermore, reliability was concerned with the ability of a measure to be consistent with theoretical and conceptual measurements. The construct validity of the items could also stress on the abstract construct of the test actually measured, regardless of test item content or other factors (Sekaran and Bougie, 2010), whereas the related construct and face validity was in reference to the degree of correlation between the measure and the criterion of voter s decision predictors as included in the hypotheses established in the structural model. All measures suggested high reliabilities, with coefficient alphas surpassing or getting closer to the acceptable level of (Hair et al., 2009) in all cases. All measures survived an exploratory factor analysis with high factor loadings and possessed unidimensionality (Tabachnick & Fidell, 2007). All in all, the measures showed a good performance and 193

214 confirmed items and indicators were conducted based on the low items-to-total correlation and cross-loadings. In sum, according to the findings from the pilot study, all measures were regarded as reliable and unidimensional, and they were adequately robust to enable further analysis and hypothesis testing in the main study. 5.3 Reliability Test Results of Actual Data GE12 (N=505) Reliability of Internet Utilisation A reliability test using Cronbach s alpha was carried out to measure the reliability level of the Internet utilisation. A total of 13 items tested the construct of Internet utilisation as a variable based on the related questions on Internet utilisation. A Cronbach s Alpha s value of more than 0.8 indicates a high level of reliability (Cohen and Cohen, 1988). Table 5.11a Reliability Statistics of Internet Utilisation Variable Cronbach s Alpha No. of Items Remark Internet Utilisation High Reliability Based on Table 5.11a, the present study confirmed a high level of reliability of Internet Utilisation with a number of item measurements was 13. This study also configured a detail reliability test for each item. It confirmed a level of high reliability for items number 1 until item number 13 (Table 5.11b). The consistency of indicators had a significant interaction as a reliable measurement of Internet Utilisation. The reliability test was useful to confirm all measurements of variable and all indicators were involved further with confirmatory factor analysis (CFA) process until hypothesis testing 194

215 Table 5.11b Item-Total Statistics of Internet Utilisation Items Cronbach s Alpha Corrected Item- Remark Total Correlation Utilisation High Reliability Utilisation High Reliability Utilisation High Reliability Utilisation High Reliability Utilisation High Reliability Utilisation High Reliability Utilisation High Reliability Utilisation High Reliability Utilisation High Reliability Utilisation High Reliability Utilisation High Reliability Utilisation High Reliability Utilisation High Reliability Reliability of Candidate Profiles Using Cronbach Alpha, the reliability test was carried out to measure the candidate profile s reliability level. There were 9 items tested as constructs of candidate profiles as a variable based on the questions related to candidate profile. The Cronbach s Alpha confirmed the high level of reliability analysis. Similar to the results obtained in the pilot test, the data value was more than 0.9, and based on Cohen and Cohen (1988), it showed a high level of reliability. Table 5.12a Reliability Statistics of Candidate Profiles Variable Cronbach s Alpha No. of Items Remark Candidate Profile High Reliability Based on Table 5.12a, the present study confirmed a high level of reliability of Candidate Profile with a number of item measurements were 9. This study also configured a detail reliability test for each item. It confirmed a high level of reliability for items number 1 until item number 9 (Table 5.12b). The consistency of indicators had a significant interaction as a 195

216 reliable measurement of Candidate Profile. The reliability test was useful to confirm all measurements of variable and all indicators were involved further with confirmatory factor analysis (CFA) process until hypothesis testing Table 5.12b Item-Total Statistics of Candidate Profiles Items Cronbach s Alpha Corrected Item- Remark Total Correlation Profile High Reliability Profile High Reliability Profile High Reliability Profile High Reliability Profile High Reliability Profile High Reliability Profile High Reliability Profile High Reliability Profile High Reliability Reliability of Party Orientation A reliability test using Cronbach s alpha was employed to measure the reliability level of Internet utilisation, with 3 dimensions tested for the construct of party orientation as a variable. A Cronbach s Alpha of more than 0.7 confirmed the high level of reliability analysis based on Cohen (1988) criteria. Table 5.13a to Table 5.16b set out the detailed results of the testing. 196

217 Table 5.13a Reliability Statistics of Party Orientation Variable Cronbach s Alpha No. of Items Remark Party Orientation High Reliability Based on Table 5.13a, the present study confirmed a high level of reliability of Party Orientation with a number of item measurements were 7. This study also configured a detail reliability test for each item. It confirmed a high reliability for items number 1 until item number 7 (Table 5.13b). The consistency of indicators had a significant interaction as a reliable measurement of Party Orientation. The reliability test was useful to confirm all measurements of variable and all indicators were involved further with confirmatory factor analysis (CFA) process until hypothesis testing (Table 5.13a and Table 5.13b). Table 5.13b Item-Total Statistics of Party Orientation Items Cronbach s Alpha Corrected Item- Remark Total Correlation Orientation High Reliability Orientation High Reliability Orientation High Reliability Orientation High Reliability Orientation High Reliability Orientation High Reliability Orientation High Reliability A. Reliability of Product Orientation Based on Table 5.14a, the present study confirmed a high level of reliability of Product Orientation with a number of item measurements were 5. This study also configured a detail reliability test for each item. It confirmed a high reliability for items number 1 until item number 5 (Table 5.14b). The consistency of indicators had a significant interaction as a reliable measurement of Product Orientation. The reliability test was useful to confirm all 197

218 measurements of variable and all indicators were involved further with confirmatory factor analysis (CFA) process until hypothesis testing and validation of the structural model. Table 5.14a Reliability Statistics of Product Orientation Variable Cronbach s Alpha No. of Items Remark Product Orientation High Reliability Table 5.14b Item-Total Statistics of Product Orientation Items Cronbach s Alpha Corrected Item- Remark Total Correlation Product High Reliability Product High Reliability Product High Reliability Product High Reliability Product High Reliability B. Sales Orientation Based on Table 5.15a, the present study confirmed a high level of reliability of Sales Orientation with a number of item measurements were 5. This study also configured a detail reliability test for each item. It confirmed a high level of reliability for items number 1 until item number 5 (Table 5.15b). The consistency of indicators had a significant interaction as a reliable measurement of Sales Orientation. The reliability test was useful to confirm all measurements of variable and all indicators were involved further with confirmatory factor analysis (CFA) process until hypothesis testing 198

219 Table 5.15a Reliability Statistics of Sales Orientation Variable Cronbach s Alpha No. of Items Remark Sales Orientation High Reliability Table 5.15b Item-Total Statistics of Sales Orientation Items Cronbach s Alpha Corrected Item- Remark Total Correlation Sales High Reliability Sales High Reliability Sales High Reliability Sales High Reliability Sales High Reliability C. Marketing Orientation Based on Table 5.16a, the present study confirmed a high level of reliability of Marketing Orientation with a number of item measurements were 5. This study also configured a detail reliability test for each items. It confirmed a high level of reliability for items number 1 until item number 5 (Table 5.16b). The consistency of indicators had a significant interaction as a reliable measurement of Marketing Orientation. The reliability test was useful to confirm all measurements of variable and all indicators were involved further with confirmatory factor analysis (CFA) process until hypothesis testing Table 5.16a Reliability Statistics of Marketing Orientation Variable Cronbach s Alpha No. of Items Remark Marketing Orientation High Reliability 199

220 Table 5.16b Item-Total Statistics of Marketing Orientation Items Cronbach s Alpha Corrected Item- Remark Total Correlation Marketing High Reliability Marketing High Reliability Marketing High Reliability Marketing High Reliability Marketing High Reliability In the reliability analysis, 5 items were tested to measure each of Sales, Product and Market as dimensions of party orientation. The outcome of the analysis validated 3 dimensions and 15 items as constructs of party orientation as hypothesised. Cronbach s Alpha for variables and also Cronbach s Alpha for items (Cronbach s Alpha if item deleted) confirmed that the dimensions were reliable for further analysis of the structural model Reliability of Political Aptitude Reliability test using Cronbach s alpha was employed to measure the reliability level of political aptitude. There were 19 items tested for the construct of political aptitude as a variable. Detailed results of the testing are set out in Table 5.17a and Table 5.17b. Table 5.17a Reliability Statistics of Political Aptitude Variable Cronbach s Alpha No. of Items Remark Political Aptitude High Reliability Based on Table 5.17a, the present study confirmed a high level of reliability of Political Aptitude with a number of item measurements were 19. This study also configured a detail reliability test for each items. It confirmed a high reliability for items number 1 until item number 19 (Table 5.17b). The consistency of indicators had a significant interaction as a reliable measurement of Political Aptitude. The reliability test was useful to confirm all 200

221 measurements of variable and all indicators were involved further with confirmatory factor analysis (CFA) process until hypothesis testing and interaction effect analysis. Table 5.17b Item-Total Statistics of Political Aptitude Items Cronbach s Alpha Corrected Item- Remark Total Correlation Aptitude High Reliability Aptitude High Reliability Aptitude High Reliability Aptitude High Reliability Aptitude High Reliability Aptitude High Reliability Aptitude High Reliability Aptitude High Reliability Aptitude High Reliability Aptitude High Reliability Aptitude High Reliability Aptitude High Reliability Aptitude High Reliability Aptitude High Reliability Aptitude High Reliability Aptitude High Reliability Aptitude High Reliability Aptitude High Reliability Aptitude High Reliability Reliability of Political Activity (Voters Like) A reliability test was done to measure the reliability level of political activities, using Cronbach s alpha. There were 10 items tested for the construct of political activity as a variable. The Cronbach s Alpha obtained confirmed the high level of reliability analysis. The value was greater than 0.8, and if we were to refer to Cohen (1988), this depicted a high level of reliability. The detailed results of the testing are established in Table 5.18a and Table 5.18b. 201

222 Table 5.18a Reliability Statistics of Activity (Voters Like) Variable Cronbach s Alpha No. of Items Remark Activity (Voters Like) High Reliability Table 5.18b Item-Total Statistics of Activity (Voters Like) Items Cronbach s Alpha Corrected Item- Remark Total Correlation Like High Reliability Like High Reliability Like High Reliability Like High Reliability Like High Reliability Like High Reliability Like High Reliability Like High Reliability Like High Reliability Like High Reliability Based on Table 5.18a, the present study confirmed a high level of reliability of Activity (Voters Like) with a number of item measurements were 10. This study also configured a detail reliability test for each item. It confirmed a high level of reliability for items number 1 until item number 10 (Table 5.18b). The consistency of indicators had a significant interaction as a reliable measurement of Activity (Voters Like). The reliability test was useful to confirm all measurements of variable and all indicators were involved further with confirmatory factor analysis (CFA) process until hypothesis testing Reliability of Voters Decision A reliability test using Cronbach s alpha was employed to measure the reliability level of voters decision as reported by the 505 respondents to the questionnaire in the study. Eleven items were tested for the construct of voters decision as a variable. The Cronbach s Alpha 202

223 values confirmed the high level of reliability: with a value greater than 0.8, based on Cohen (1988) criteria, this represented a high level of reliability. Table 5.19a contains the detailed results of the testing of variable Voters Decision. Table 5.19a Reliability Statistics of Voters Decision Variable Cronbach s Alpha No. of Items Remark Voters Decision High Reliability Based on Table 5.19a, the present study confirmed a high level of reliability of Voters Decision with a number of item measurements were 11. This study also configured a detail reliability test for each item. It confirmed a high level of reliability for items number 1 until item number 11 (Table 5.19b). The consistency of indicators had a significant interaction as a reliable measurement of Voters Decision. The reliability test was useful to confirm all measurements of variable and all indicators were involved further with confirmatory factor analysis (CFA) process until hypothesis testing (Table 5.19b). 203

224 Table 5.19b Item-Total Statistics of Voters Decision Items Cronbach s Alpha Corrected Item- Remark Total Correlation Decision High Reliability Decision High Reliability Decision High Reliability Decision High Reliability Decision High Reliability Decision High Reliability Decision High Reliability Decision High Reliability Decision High Reliability Decision High Reliability Decision High Reliability 5.4 Confirmatory Factor Analysis (CFA) of Variables (GE12) The number of factors and the loadings of items on them must be determined by the CFA, to conform to what was expected based on the pre-established theory on scale assessment (Hair et al., 2010). The Structural Equation Modeling techniques were helpful as they performed the CFA. The AMOS 21.0 version computer program served to investigate if the proposed factor solutions fit the data (Sentosa et al., 2012). No cross-loadings were permitted. All scale items used for the CFAs were outlined in this section and several measures were used in SEM to evaluate the model fit. The supposed direct causal effect of the latent variable on the observed measure from a factor to an indicator is depicted by a single arrow-head line, which points from a factor to an indicator in a standard CFA model. Known as factor loadings, the statistical estimates were interpreted as regression coefficients, and they were then tested in the hypothesised model (Byrne, 2001). 204

225 The Confirmatory Factor Analysis (CFA) was later on conducted to configure and confirm the variables and detailed measurements according to the hypotheses. There were 5 first-order CFA of variables (Internet Utilisation, Candidate Profiles, Political Aptitude and Voters Likes) and one second-order CFA (Party Orientation) with three dimensions: Party Orientation, Marketing Orientation and Sales Orientation. First order CFA also validated Voters` Decision as an endogenous variable. The goodness of fit indexes (Chi-Square, Degree of Freedom, Ratio, P-Value, GFI, TLI and RMSEA) were adopted as indicators to confirm the significant model that had valid standardised regression weights (factor loading > 0.5) and low level of standard error (< 10%) CFA of Exogenous Variables A. CFA of Internet Utilisation Internet Utilisation as an exogenous variable was measured and validated using confirmatory factor analysis (CFA). A total of 13 items were validated using 505 respondents as the sample (Figure 5.1a). The standardised regression weight (Figure 5.1b and Table 5.20) showed the convergent reliability of 7 valid items (Factor loading more than 0.5). The CFA also confirmed the significant indexes of the CFA model (Figure 5.1b). 1 st order CFA of Internet Utilisation confirmed the achievement of convergent validity and achievement of model fit (Ratio less than 2; GFI and TLI > 0.9 and RMSEA < 0.08). 205

226 Figure 5.1a CFA of Internet Utilisation 206

227 Figure 5.1b CFA of Internet Utilisation 207

228 Table 5.20 Standardised Regression Weights of Internet Utilisation Items Variable Estimate S.E. C.R. P Remark Utilisation2 <--- Internet Utilisation Significant Utilisation3 <--- Internet Utilisation Significant Utilisation4 <--- Internet Utilisation Significant Utilisation5 <--- Internet Utilisation Significant Utilisation7 <--- Internet Utilisation Significant Utilisation9 <--- Internet Utilisation Significant Utilisation10 <--- Internet Utilisation Significant B. CFA of Candidate Profiles The CFA of candidate profiles were conducted by constraining the original 9 items of the endogenous constructs (Figure 5.2a). Closely referring to the modification index and error elimination, the CFA of candidate profiles verified that there were 7 valid items in the model development. The CFA of the candidate profiles constructs yielded a relatively good fit as shown by the goodness-of-fit. The measurement model showed that it had a good fit with the data (Table 5.21) based on a few assessment criteria such as GFI, CFI, TLI, and RMSEA (Bagozzi & Yi, 1988). Figure 5.2b highlights the goodness of fit of the CFA candidate profiles. Table 5.21 Standardised Regression Weights of Candidate Profiles Items Variable Estimate S.E. C.R. P Remark Profile1 <--- Candidate Profiles Significant Profile2 <--- Candidate Profiles Significant Profile3 <--- Candidate Profiles Significant Profile4 <--- Candidate Profiles Significant Profile5 <--- Candidate Profiles Significant Profile7 <--- Candidate Profiles Significant Profile8 <--- Candidate Profiles Significant 208

229 Figure 5.2a CFA of Candidate Profiles 209

230 Figure 5.2b CFA of Candidate Profiles 210

231 C. CFA of Party Orientation The 2nd order analysis of the party orientation category, an exogenous construct of the measurement model, was carried out to evaluate the measures unidimensionality (Byrne, 2001). As hypothesised, there were 3 dimensions of party orientation; Product, Sales and Marketing. Figure 5.3a shows the original model of the measurement model. Figure 5.3a CFA of Party Orientation 211

232 Figure 5.3b CFA of Party Orientation 212

233 The 2nd order CFA of party orientation was conducted by constraining the original 15 items of the endogenous constructs (Figure 5.3b). With regard to the modification index and error elimination, the CFA of party orientation made a confirmation that there were 12 valid items in the model development. The CFA of candidate profiles constructs produced a relatively good fit as shown by the goodness-of-fit indices such as CMIN/DF ratio (<2); P-value (>0.05); the Goodness-of-Fit Index (GFI) of >.90; and RMSEA values less than.08 (<.08), (Table 5.22) Figure 5.3b confirmed the significant construct of Party Orientation. Using 2nd order analysis CFA, three dimensions of Party Orientation were validated: Product, sales and marketing orientation. 2nd order CFA showed that sales orientation had the highest loading (factor loading = 0.99); marketing orientation also had a high level of loading (factor loading = 0.96), followed by product orientation with the lowest loading (factor loading = 0.86) within the construct as a whole. Validation identified four significant items for each dimension. Convergent Reliability was achieved with a factor loading value of more than 0.5. Details of the loadings are shown in Table

234 Table 5.22 Standardised Regression Weights of Party Orientation (2 nd Order CFA) Items Variable Estimate S.E. C.R. P Label Sales <--- Party Orientation Significant Marketing <--- Party Orientation Significant Product <--- Party Orientation Significant Product2 <--- Product Significant Product3 <--- Product Significant Product4 <--- Product Significant Product5 <--- Product Significant Sales2 <--- Sales Significant Sales3 <--- Sales Significant Sales4 <--- Sales Significant Sales5 <--- Sales Significant Marketing5 <--- Marketing Significant Marketing3 <--- Marketing Significant Marketing2 <--- Marketing Significant Marketing1 <--- Marketing Significant D. CFA of Political Aptitude The CFA of political aptitude was conducted by way of constraining the original 19 items of the endogenous constructs (Figure 5.4a). Leaning on the modification index and error elimination, the CFA of political aptitude confirmed that there were 5 valid items in the development of the model. The CFA of political aptitude constructs (Table 5.23) showed a relatively good fit, suggested by the goodness-of-fit indices such as CMIN/df ratio (<2); P- value (>0.05); the Goodness-of-Fit Index (GFI) of >.90; and RMSEA values less than.08 (<.08). Furthermore, Figure 5.4b also confirmed 5 items as valid measurements of political aptitude (with a factor loading greater than 0.5). 214

235 Figure 5.4a CFA of Political Aptitude 215

236 Figure 5.4b CFA of Political Aptitude 216

237 Table 5.23 Standardised Regression Weights of Political Aptitude Items Variable Estimate S.E. C.R. P Label Aptitude5 <--- Political Aptitude Significant Aptitude6 <--- Political Aptitude Significant Aptitude7 <--- Political Aptitude Significant Aptitude8 <--- Political Aptitude Significant Aptitude9 <--- Political Aptitude Significant E. CFA of Political Activities The CFA of political activities was conducted by constraining the original 10 items of the endogenous constructs (Figure 5.5a). The CFA of political activities confirmed that there were 5 valid items in the development of the model, based on both the modification index and error elimination (Figure 5.5b and Table 5.24). The CFA of political activities constructs had a relatively good fit, as suggested by the goodness-of-fit indices such as CMIN/df ratio (<2); P-value (>0.05); the Goodness-of-Fit Index (GFI) of >.90; and RMSEA of values less than.08 (<.08). Table 5.24 Standardised Regression Weights of Voters Like Items Variable Estimate S.E. C.R. P Remark Like1 <--- Voters Like Significant Like2 <--- Voters Like Significant Like3 <--- Voters Like Significant Like4 <--- Voters Like Significant Like6 <--- Voters Like Significant 217

238 Figure 5.5a CFA of Political Activities (Voters Like) 218

239 Figure 5.5b CFA of Political Activities (Voters Like) 219

240 5.4.2 CFA of Endogenous Variable (Voters Decision) The CFA of voters decision was conducted by constraining the original11 items of the endogenous constructs (Figure 5.6a). Based on the modification index and error elimination, the CFA of voter s decision confirmed that there were 5 valid items in the development of the model (Figure 5.6b and Table 5.25). The CFA of voter s decision constructs produced a relatively good fit as indicated by the goodness-of-fit indices such as CMIN/DF ratio (<2); P- value (>0.05); the Goodness-of-Fit Index (GFI) of >.90; and RMSEA of values less than.08 (<.08). Table 5.25 Standardised Regression Weights of Voters Decision Items Variable Estimate S.E. C.R. P Label Decision5 <--- Voters Decision Significant Decision7 <--- Voters Decision Significant Decision8 <--- Voters Decision Significant Decision9 <--- Voters Decision Significant Decision11 <--- Voters Decision Significant 220

241 Figure 5.6a CFA of Voters Decision 221

242 Figure 5.6b CFA of Voters Decision 222

243 5.5 Measurement Model of Exogenous Variables (Multicollinearity Testing) Correlation testing in the structural model was employed using the measurement model of exogenous variables. The objective of the testing was to identify any multicollinearity among predictors in the structural model. However, the test confirmed the absence of multicollinearity among exogenous variables (correlation < 0.9) or significant difference among independent variables in the hypothesised, generated and re-specified model (Figure 5.7). e2 e4 e5 e7 e23 e22 e21 e20 SAL5.95 SAL SAL3.95 SAL2 R01 Sales R e41 LIKE1.96 e40 LIKE2.96 e38 e37 LIKE6.96 LIKE4.96 Political Activities CP CP4 CP Candidate Profiles CP8 e26 e25 PRO PRO3.95 Product.21 Party Orientation e24 PRO4 e31 e30 e29 e28 MA1 MA2 MA3 MA R03 Market.27 e36 AP9.94 e35 AP e34 AP7.95 e33 AP6 Political Aptitude Internet Utility -.04 Internet Utilisation INT2 INT7 INT9 e14 e10 e9 Figure 5.7 Measurement Model of Exogenous Variables 223

244 5.6 Goodness of Fit Indexes, Composite Reliability (CR) and Average Variance Extracted (AVE) as Discriminant Validity Testing Next, the Chi-square/degrees of freedom of the Re-specified Model had shown a goodness-offit of the model. Jöreskog and Sörbom (1993) proposed an appropriate platform for selecting the hypothesised models by deliberating on the model parsimony and fit. The use of three measures, Akaike s Information Criterion (AIC), Expected Cross-Validation Index (ECVI), and Consistent Akaike s Information Criterion (CAIC), which were the functions of model chi-square and degree of freedom, were also recommended (Akaike, 1987). The model with the smallest value of those three measures should be opted for. To add to this, the parsimonious normed fit index (PNFI) proved to be another useful measure to ascertain the model parsimony with higher values that reflect greater model parsimony (Hair et al., 2009). Furthermore, the study also confirmed the Discriminant Validity of measurements using composite reliability (CR) and average variance extracted (AVE) (Table 5.26). The test for the average variance extracted (AVE) on the constructs demonstrated that the AVE exceeded the square of the structural link between the three constructs, offering support to the discriminant validity. The fully standardised parameters (factor loadings), composite reliability and AVE of both exogenous and endogenous constructs can be seen in Table For each latent variable, Fornell and Larcker s (1981) composite reliability score was calculated. Composite reliability is a means of looking into the internal consistency of the items of a latent variable (Chin, 1988). In the present work, all political marketing factors had shown a composite reliability score of above 0.70, suggesting adequate reliability. 224

245 Table 5.26 Composite Reliability and Average Variance Extracted from Measurements Items Estimate R 2 error var εj Composite reliability Variance Extracted CP CP CP CP Total INT INT INT Total SAL SAL SAL SAL Total PRO PRO PRO Total MA MA MA MA Total AP AP AP AP Total LIKE LIKE LIKE LIKE Total

246 The AVE measures the amount of variance captured by the construct related to the amount of variance following measurement error (Fornell & Larcker, 1981). The AVE can be defined as a measure of reliability for a construct, and one recommendation is for the AVE to be greater than 0.50, and this means that the construct captures more variance in the items than the measurement error (Chin, 1988b). Note: Table 5.27 Composite Reliability and Variance Extracted of Variables Constructs Composite Reliability (CR) b Variance Extracted (VE) c Exogenous Construct Candidate Profiles Internet Utilisation Party Orientation Political Aptitude Political Activities Endogenous Construct Voters Decision Moderating Construct Facebook Time Facebook Friends a Completely standardised parameter. Square multiple correlations for each measure can be obtained by squaring the completely standardised parameter for the item (factor loadings) b Construct reliability computed as ( Σ λ ) 2 / [ (Σ λ ) 2 + Σ var (ε) ] c Average variance extracted, which is the proportion of variance in the construct that is not due to measurement error (Fornell & Larcker, 1981) Table 5.26 and Table 5.27 show that the AVE for each construct is higher than There is an indication here that each construct account for 50% or more of the items variance. The AVE statistics can also assess the discriminant validity in two ways- by calculating the square root of the AVE statistics and comparing them with the correlations among the latent variables (Zahir & Sentosa, 2013). This was a test of whether more variance is shared 226

247 between the latent variable and it was set of items than with other latent variables depicted by a different set of items of political marketing factors and voters decisions. Table 5.27 also shows that the square roots of the AVEs are all larger than their respective correlations, highlighting the fact that the discriminant validity is adequate. To confirm the discriminant validity between constructs, AVE should be greater than the squared correlation between a construct and any other construct in the model. The square root of the AVE was presented at the diagonal of the correlation matrix to enable it to be compared with the correlation coefficient between constructs (Zahir & Sentosa, 2013). Adequate discriminant validity was undoubted, as the diagonal elements are greater than the off-diagonal elements in the analogous rows and columns. The study further confirmed that there was a discriminant validity of the final measurements of the re-specified model using the procedures outlined by Fornell and Laker (1981). Discriminant validity testing confirmed the correlations between the latent variables and the average variance extracted (AVE) of each construct as seen on the diagonal. Fornell and Lacker (1981) asserted that the squared correlation between constructs must be less than the average variance extracted (AVE) of each underlying construct as for the constructs to have their own discriminant validity. 5.7 Model Validation using Current Election Data (GE13) Confirmatory Factor Analysis (CFA) of Measurements Confirmatory Factor Analysis (CFA) was conducted to configure and confirm the variables and detailed measurements as hypothesised. Using data from the latest election (GE13), five 1 st order CFA variables (Internet Utilisation, Candidate Profiles, Political Aptitude and Voters Like) and one 2 nd order CFA (Party Orientation) with three dimensions: Party Orientation, 227

248 Marketing Orientation and Sales Orientation were investigated. 1 st order CFA also validated Voters Decision as an endogenous variable. The goodness of fit indexes (Chi-Square, Degree of Freedom, Ratio, P-Value, GFI, TLI and RMSEA) were employed as indicators to confirm the significant model with valid standardised regression weights (factor loading > 0.5) and low level of standard error (< 10%) CFA of Exogenous Variables A. CFA of Internet Utilisation Internet Utilisation as an exogenous variable was measured and validated using confirmatory factor analysis (CFA). 13 items were validated using 505 respondents as the sample (Figure 5.8a). The standardised regression weights show the convergent reliability of 10 valid items (factor loading more than 0.5). The CFA also confirmed the significant indexes of the CFA model (Figure 5.8b). Figure 5.8b and Table 5.28 confirmed the validation of Internet Utilisation as one of the political marketing factors. Using data from a sample of 150 respondents who voted in the GE13, the present study succeeded in validating 9 items of Internet utilisation (Table 5.28) Table 5.28 Standardised Regression Weights of Internet Utilisation Items Variable Estimate S.E. C.R. P Remark Utilisation2 <--- Internet Utilisation Significant Utilisation3 <--- Internet Utilisation Significant Utilisation4 <--- Internet Utilisation Significant Utilisation5 <--- Internet Utilisation Significant Utilisation6 <--- Internet Utilisation Significant Utilisation7 <--- Internet Utilisation Significant Utilisation8 <--- Internet Utilisation Significant Utilisation9 <--- Internet Utilisation Significant Utilisation10 <--- Internet Utilisation Significant 228

249 Figure 5.8a CFA of Internet Utilisation 229

250 Figure 5.8b CFA of Internet Utilisation 230

251 B. CFA of Candidate Profiles The CFA of candidate profiles was conducted by means of constraining the original 9 items of the endogenous constructs based on the 9 questions related to the candidate profiles (Figure 5.9). The CFA of candidate profiles confirmed that there were 9 valid items remaining in the development of the model (Byrne, 2001) based on the modification index and error elimination (Table 5.29). The CFA of candidate profiles constructs demonstrated a relatively good fit as indicated by goodness-of-fit indices such as CMIN/df ratio (<2); P-value (>0.05); the Goodness of Fit Index (GFI) of >.90; and RMSEA values which were less than.08 (<.08). Table 5.29 Standardised Regression Weights of Candidate Profiles Items Variable Estimate S.E. C.R. P Remark Profile1 <--- Candidate Profiles Significant Profile2 <--- Candidate Profiles Significant Profile3 <--- Candidate Profiles Significant Profile4 <--- Candidate Profiles Significant Profile5 <--- Candidate Profiles Significant Profile6 <--- Candidate Profiles Significant Profile7 <--- Candidate Profiles Significant Profile8 <--- Candidate Profiles Significant Profile9 <--- Candidate Profiles Significant 231

252 Figure 5.9 CFA of Candidate Profiles 232

253 C. CFA of Party Orientation The 2nd order CFA of party orientation for GE13 was conducted by constraining the original 15 items of the endogenous constructs (Figure 5.10a). With the modification index and error elimination serving as the basis, the CFA of party orientation further verified that there were 13 valid items in the development of the model (Figure 5.10b). The 2nd order CFA of the party orientation construct yielded a relatively good fit as shown by the goodness-of-fit indices such as CMIN/df ratio (<2); P-value (>0.05); the Goodness-of-Fit Index (GFI) of >.90; and RMSEA values of less than.08 (<.08) (Table 5.30). Table 5.30 Standardised Regression Weights of Party Orientation (2ndOrder CFA) Items Variable Estimate S.E. C.R. P Label Sales <--- Party Orientation Significant Marketing <--- Party Orientation Significant Product <--- Party Orientation Significant Product2 <--- Product Significant Product3 <--- Product Significant Product4 <--- Product Significant Product5 <--- Product Significant Sales2 <--- Sales Significant Sales3 <--- Sales Significant Sales4 <--- Sales Significant Sales5 <--- Sales Significant Marketing5 <--- Marketing Significant Marketing4 <--- Marketing Significant Marketing3 <--- Marketing Significant Marketing2 <--- Marketing Significant Marketing1 <--- Marketing Significant 233

254 Figure 5.10a CFA of Party Orientation 234

255 Figure 5.10b CFA of Party Orientation for GE13 235

256 D. CFA of Political Aptitude The CFA of political aptitude was conducted by constraining the original 19 items of the endogenous constructs (Figure 5.11a). Again, leaning on the modification index and error elimination, the CFA of political aptitude confirmed that there were 8 valid items in the development of the model (Figure 5.11b and Table 5.31). Table 5.31 Standardised Regression Weights of Political Aptitude Items Variable Estimate S.E. C.R. P Label Aptitude5 <--- Political Aptitude Significant Aptitude6 <--- Political Aptitude Significant Aptitude7 <--- Political Aptitude Significant Aptitude8 <--- Political Aptitude Significant Aptitude9 <--- Political Aptitude Significant Aptitude10 <--- Political Aptitude Significant Aptitude11 <--- Political Aptitude Significant Aptitude12 <--- Political Aptitude Significant 236

257 Figure 5.11a CFA of Political Aptitude 237

258 Figure 5.11b CFA of Political Aptitude 238

259 E. CFA of Political Activities The CFA of political activities was conducted by constraining the original 10 items of the endogenous constructs (Figure 5.12a). Leaning on the modification index and error elimination, the CFA of political activities confirmed that there were 8 valid items in the development of the model (Figure 5.12b and Table 5.32). The CFA of the candidate profiles construct showed a relatively good fit as highlighted by the goodness-of-fit indices such as CMIN/df ratio (<2); P-value (>0.05); the Goodness-of-Fit Index (GFI) of >.90; and RMSEA values of less than.08 (<.08). Table 5.32 Standardised Regression Weights of Voters Like Items Variable Estimate S.E. C.R. P Remark Like1 <--- Voters Like Significant Like2 <--- Voters Like Significant Like3 <--- Voters Like Significant Like4 <--- Voters Like Significant Like5 <--- Voters Like Significant Like6 <--- Voters Like Significant Like8 <--- Voters Like Significant Like9 <--- Voters Like Significant 239

260 Figure 5.12a CFA of Political Activities (Voters Like) 240

261 Figure 5.12b CFA of Political Activities (Voters Like) 241

262 5.7.3 CFA of Endogenous Variable (Voters Decision) The CFA of voters decision was conducted by constraining the original11 items of the endogenous constructs (Figure 5.13a). Referring to the modification index and error elimination, the CFA of voters decision confirmed that there were 9 valid items in the development of the model (Figure 5.13b). The CFA of the candidate profiles construct showed a relatively good fit as indicated by the goodness-of-fit indices such as CMIN/DF ratio (<2); P-value (>0.05); the Goodness-of-Fit Index (GFI) of >.90; and RMSEA values of less than.08 (<.08), (Table 5.33). Table 5.33 Standardised Regression Weight of Voters Decision Items Variable Estimate S.E. C.R. P Label Decision3 <--- Voters Decision Significant Decision4 <--- Voters Decision Significant Decision5 <--- Voters Decision Significant Decision6 <--- Voters Decision Significant Decision7 <--- Voters Decision Significant Decision8 <--- Voters Decision Significant Decision9 <--- Voters Decision Significant 242

263 Figure 5.13a CFA of Voters Decision 243

264 Figure 5.13b CFA of Voters Decision 244