Brand Communication on Social Networking Sites Freya De Keyzer

Home , Highlight (application), Koo (social network), LinkedIn, Mention (blogging)

Proefschrift voorgelegd tot het behalen van de graad van doctor in de Toegepaste Economische Wetenschappen aan de Universiteit Antwerpen:

Departement Marketing

Faculteit Bedrijfswetenschappen en Economie

Universiteit Antwerpen

Promotoren: Prof. dr. Nathalie Dens Prof. dr. Patrick De Pelsmacker

Antwerpen, 2019

Doctoral Jury

Prof. dr. Patrick De Pelsmacker (supervisor)

University of Antwerp & Ghent University

Prof. dr. Nathalie Dens (supervisor)

University of Antwerp & Antwerp Management School

Prof. dr. Karolien Poels (chair)

University of Antwerp

Prof. dr. Ingrid Moons (secretary)

University of Antwerp

Prof. dr. Tim Smits

KU Leuven

Prof. dr. Guda van Noort (member of the doctoral committee)

University of Amsterdam

Prof. dr. Kim Willems

Vrije Universiteit Brussel

Acknowledgments

In 2012 studeerde ik af als Master in de Communicatiewetenschappen en ik moet eerlijk bekennen dat de 22-jarige ik absoluut geen idee had wat aan te vangen met dat diploma. Het heeft me dan ook meer dan een half jaar én veel gesprekken met mensen rondom mij gekost om te beseffen dat ik onderzoek doen misschien toch wel fijn vond en dat ik daar misschien toch maar iets mee wilde gaan doen. Enkele maanden later stapte ik, letterlijk ziek van de zenuwen, een voor mij nieuwe en totaal ongekende faculteit binnen. Tijdens een infosessie in één van mijn eerste weken als mandaatassistent kreeg ik slaagpercentages te zien van assistenten die een doctoraat aanvatten... Ondanks het feit dat ik de percentages niet meer van buiten ken, herinner ik me wel dat die cijfers een grote impact hadden: “Zal me dit wel lukken?”

Ik moet dan ook bekennen dat ik vandaag, zo’n zes jaar later, trots ben om die statistiek, in de spreekwoordelijke zin van het woord, verslagen te hebben. Ik vertel aan iedereen die het me vraagt dat ik het een absoluut privilege vind om een doctoraat te mogen doen, maar desondanks is het absoluut geen gemakkelijke weg geweest is met enkele pieken, maar toch ook met behoorlijk wat dalen. Zeker aan het begin van een doctoraat maak je een ongezien steile leercurve mee op vlak van onderzoek, maar ook op vlak van onderwijs. Ik heb immers de afgelopen jaren mijn tijd verdeeld over onderzoek en onderwijs. Het zoeken van een mooie balans tussen die twee is zeker niet altijd evident geweest en ik moet dan ook toegeven dat opgeven één van de dingen is die doorheen het traject toch wel enkele keren door mijn hoofd heeft gespeeld. Bovendien zorgde mijn FOMO (fear-of-missing-out) er voor dat ik me steeds head-first in allerlei projecten en opdrachten stortte (en dat nog steeds doe), gewoon omdat het interessant is. Het zijn de mensen rondom me die mij gesteund hebben om door te gaan en die me ondersteund hebben in allerlei dingen die ik wilde ondernemen, die het verdienen om hier toch heel expliciet in de bloemetjes te worden gezet.

Nathalie, bedankt voor je onuitputtelijke inzet en toewijding. Jouw feedback op manuscripten en presentaties was bij momenten erg pittig en ik heb daar mee moeten leren omgaan, maar de afgelopen zes jaar heb ik geleerd dat jouw doel vooral is om ons werk, en bij uitbreiding mij als onderzoeker, beter te maken. Als promotor was jij diegene die de puntjes op de i zette, de details nalas en toch telkens nog ergens een foutje opmerkte. Ik heb enorm veel van jou geleerd. Bedankt!

Patrick, bedankt om je expertise en ervaring met mij te willen delen. Ik weet dat je een dankwoord als dit een soort verplicht nummertje vindt, toch hoop ik je te kunnen overtuigen van mijn oprechte dank. i Als promotor was jij de man van de grote lijnen: “de flow moet goed zitten” en “je moet zorgen dat je de lezer bij de hand neemt” zijn maar enkele van opmerkingen in dat kader die me bij blijven. Ook van jou heb ik erg veel geleerd. Bedankt!

Speciale dank gaat ook uit naar de collega’s van het Marketing departement. Kristien, bedankt voor je steun en voor de véle babbels. Door jou ben ik vaak een uur later thuis gekomen dan aangekondigd omdat ik snel nog even iets wilde zeggen. Bram, samen met Kristien, zorgde je ervoor dat ik al na 1 jaar newbie af was. Bedankt voor de gesprekken en de gedeelde smart in de laatste weken van mijn doctoraat. Ingrid, bedankt voor de babbeltjes, voor het willen zetelen in mijn jury, voor het meedenken aan het FWO project dat we binnenkort zullen omvormen in een post-doc aanvraag en voor het mee willen denken aan mijn toekomst. Annouk, bedankt voor je steun en om af en toe te vragen “En meiske, lukt het een beetje?” Yana, thank you for your constructive feedback over the past six years. I must say that I do look up to you and I am looking forward to brainstorming on a new project that combines our research topics (we really need to do that!). Cristian and Ana, you are mentioned in one breath (I’m sorry for that). Thank you for listening to me complaining, for your unconditional support and for the one - or maybe two - glasses of wine we had during the past conferences.

Bedankt ook aan de leden van mijn jury. Karolien, bedankt voor jouw jaarlijkse constructieve feedback, die dit doctoraat ongetwijfeld sterker heeft gemaakt. Guda, bedankt voor jouw positieve bijdrage tijdens onze jaarlijkse vergaderingen en je steun tijdens congrespresentaties. Bedankt om me in 2016 samen met Sanne te willen begeleiden toen ik bij jou op onderzoeksverblijf kwam en om me aan te moedigen onze studie op verschillende congressen te presenteren. Tim, bedankt voor de interessante vragen en gesprekken tijdens congressen en bedankt om aan mij te denken voor de invulling van jouw vervanging.

Kim, bedankt om de tijd te nemen om in mijn jury te zetelen.

Bedankt aan mijn twee grote broers. Marijn, je bent er altijd als Benjamin, Eloise of ik je nodig hebben en ondanks dat je een man van weinig woorden bent, weet ik dat je trots bent. Robin, bedankt voor het werk aan het eerste stimulusmateriaal. Bedankt om af en toe te vragen wanneer ik nu eindelijk eens zou beginnen werken, dat houdt een mens wel met de voetjes op de grond.

Bedankt papa, om me zes jaar geleden te overtuigen om toch maar de solliciteren op de vacature aan de faculteit (toen nog) Toegepaste Economische Wetenschappen, want je weet maar nooit. Bedankt ook voor de vele (nachtelijke) babbels die ervoor gezorgd hebben dat ik de kritische zin kon ontwikkelen die ik nu elke dag nodig heb. Bedankt mama, voor alle zorgen, voor je luisterende oor doorheen de afgelopen jaren, voor de afleiding door af en toe er samen even op uit te trekken. Bedankt aan beiden om me de kans te geven verder te studeren en om me daarin te steunen, ook al maakte ik soms keuzes die niet voor de hand lagen of die achteraf niet de juiste bleken. Bedankt om er tijdens de afgelopen maanden extra te zijn voor mij, voor Benjamin en voor Eloise.

Dat brengt me bij de twee meest belangrijke personen die ik wil bedanken. Benjamin, bedankt voor je geduld, om af en toe gewoon te luisteren als ik weer maar eens begon te razen, gewoonweg om dit avontuur samen met mij te doorstaan. Het was ‘a hell of ride’. We kozen, zoals ik dat meestal doe, niet voor de makkelijkste weg: we kochten een huis dat we samen moesten verbouwen en beslisten, tegen beter weten in, om een grote renovatie te ondernemen in de laatste maanden van mijn doctoraat.

Bedankt om de last daarvan op jouw schouders te nemen. Bedankt ook voor de vele, vaak ook nachtelijke, uren waarin we samen aan stimulusmateriaal werkten; zonder jou zou dit doctoraat niet tot stand zijn gekomen! Eloise, je kan het nu nog niet begrijpen, maar misschien lees je het ooit nog wel: de afgelopen twee jaar zorgde jij voor de grootste afleiding. Dat was soms moeilijk en soms broodnodig.

Bedankt voor wie je bent.

Table of Contents

Introduction ...... 1 Context and Relevance ...... 2 Structure of the dissertation ...... 7 ii Theoretical background ...... 14 Let’s Get Personal: Which Elements Elicit Perceived Personalization in Advertising ...... 21 Abstract ...... 22 Introduction ...... 23 Theoretical Background ...... 25 Method ...... 41 Analyses and Results ...... 47 Discussion ...... 53 Managerial implications ...... 54 Limitations and Suggestions for Future Research ...... 55 Is this for me? How Consumers Respond to Personalized Advertising on Social Network Sites, ...... 59 Abstract ...... 60 Introduction ...... 61 Literature Review ...... 63 Study 1 ...... 67 Study 2 ...... 71 General Discussion ...... 74 Suggestions for future research ...... 78 How and When Personalized Advertising Leads to Brand Attitude, Click and WOM Intention ...... 81 Abstract ...... 82 Introduction ...... 83 Study 1 ...... 85 Study 2 ...... 92 General Discussion and Conclusion ...... 111 Managerial implications ...... 114 Limitations and Future Research ...... 115 Going too far? How Consumers Respond to Personalized Advertising from Different Sources, ...... 117 Abstract ...... 118 Introduction ...... 119 Personalized Advertising: Explaining Effects Using the Privacy Calculus Theory ...... 121 The Benefit Side of Personalized Advertising: Perceived Relevance ...... 123 Investigating Boundaries: The moderating effect of source type ...... 124 Method ...... 127 Results ...... 130 Discussion ...... 136 Theoretical and Practical Implications ...... 137

Agenda for Future Research ...... 139 Don’t be so emotional! How tone of voice and service type affect the relationship between message valence and consumer responses to WOM in social media, ...... 143 Abstract ...... 144 Introduction ...... 145 Literature review and hypotheses ...... 147 Study design and procedure ...... 152 Results ...... 155 Discussion and conclusion ...... 157 Managerial implications ...... 158 Limitations and suggestions for future research ...... 159 The Impact of Relational Characteristics on Consumer Responses to Word-of-Mouth on Social Networking Sites, ...... 161 Abstract ...... 162 Introduction ...... 163 Literature Review and Hypotheses ...... 166 Empirical Study ...... 175 Analyses and Results ...... 183 Discussion ...... 186 Conclusion ...... 187 Limitations and Suggestions for Future Research ...... 189 General Conclusion ...... 193 Main Findings ...... 194 Theoretical contributions ...... 198 Managerial contributions ...... 202 Limitations and directions for future research ...... 204 References ...... 209 Appendix ...... 231 Nederlandse Samenvatting...... 317

List of Tables

Table 2.1. Overview of experimental studies manipulating personalization ...... 26 Table 2.2. Orthoplan ...... 43 Table 2.3. Manipulations per attribute...... 44 Table 2.4. Measures ...... 45 Table 2.5. Conjoint analysis for perceived personalization: pooled across products ...... 48 Table 2.6. Conjoint analysis for perceived personalization: products pooled across product iii characteristics ...... 50 Table 2.7. Conjoint analysis for perceived personalization: pooled across demographic characteristics ...... 52 Table 3.1. Unstandardized Regression Weights (Study 1) ...... 70 Table 3.2. Unstandardized Regression Weights (Study 2) ...... 72 Table 4.1. Measures (Study 1) ...... 88 Table 4.2. Unstandardized regression weights with perceived personal relevance and perceived intrusiveness as mediators (Study 1) ...... 89 Table 4.3. Pretest: average perceived personalization per advertisement ...... 99 Table 4.4. Overview of manipulations per personalization element (translated from Dutch) ...... 99 Table 4.5. Overview of manipulations per condition ...... 100 Table 4.6. Measures (Study 2) ...... 101 Table 4.7. Unstandardized regression weights with perceived personal relevance and perceived intrusiveness as mediators (Study 2) ...... 103 Table 4.8. Unstandardized regression weights with attitude toward the social networking site as moderator (Study 2) ...... 105 Table 4.9. Unstandardized regression weights with perceived privacy protection by the social network site as moderator (Study 2) ...... 105 Table 4.10. Measurement model (Study 2) ...... 108 Table 4.11. Square root of average variance extracted and correlation per construct ...... 109 Table 4.12. Unstandardized regression coefficients for brand attitude, positive word-of-mouth intention and click intention (Study 2: Extending the Mechanism)...... 109 Table 4.13. Overview of the hypotheses ...... 112 Table 5.1. Measures Used in the Study ...... 129 Table 5.2. Square root of average variance extracted and correlations per factor ...... 131 Table 5.3. Path coefficients ...... 132 Table 6.1. Construct items ...... 154 Table 6.2. Unstandardized Regression Weights ...... 155 Table 7.1. Measures ...... 180 Table 7.2. Square root of average variance extracted and correlations per factor ...... 182 Table 7.3. Standardized Regression Weights for Movie Watch Intention with only interpersonal relational characteristics ...... 183 Table 7.4. Standardized Regression Weights for WOM intention with only interpersonal relational characteristics ...... 184 Table 7.5. Standardized Regression Weights for Movie Watch Intention with only person-to-site relational characteristics ...... 184 Table 7.6. Standardized Regression Weights for WOM intention with only person-to-site relational characteristics ...... 185 Table 7.7. Standardized Regression Weights for Movie Watch Intention ...... 185 Table 7.8. Standardized Regression Weights for WOM intention ...... 186

List of Figures

Figure 1.1. Overview of dissertation...... 13 Figure 2.1. Conceptual framework ...... 24 Figure 2.2. Facebook advertisement using location by Coca-Cola Bottling Co. Consolidated ...... 30 Figure 2.3. Facebook advertisement using age, gender and location from High On Life...... 32 Figure 2.4. Facebook advertisement using a life-event by Walmart ...... 34 Figure 2.5. Facebook advertisement using a life-event by Pampers ...... 34 Figure 2.6. Facebook advertisement using interests by Krachttraining Voor Vrouwen ...... 35 Figure 2.7. Facebook advertisement using friend referrals by Centerparcs ...... 37 iv Figure 3.1. Conceptual framework ...... 62 Figure 3.2. Conditional direct effect of perceived personalization on click intention (Study 2) ...... 73 Figure 4.1. Conceptual framework Study 1 ...... 87 Figure 4.2. Conceptual framework Study 2 (Investigating Boundaries) ...... 92 Figure 4.3. Conceptual framework Study 2: Extending the Mechanism ...... 95 Figure 4.4. Conditional indirect effects of perceived personalization via perceived relevance on click intention at different levels of perceived privacy protection by the SNS ...... 107 Figure 5.1. Conceptual Framework ...... 127 Figure 5.2. Bootstrap intervals of specific indirect effects of the personalization elements ...... 133 Figure 5.3. Bootstrap intervals of specific indirect effects of personalization for each source type .... 135 Figure 6.1. Conceptual framework ...... 148 Figure 6.2. Two-way interaction plot with word-of-mouth intention as dependent variable ...... 156 Figure 7.1. Conceptual Framework ...... 175

List of Appendices

Appendix 1 – Stimulus Materials ...... 232 1.A. Stimuli used in Chapter 2 ...... 232 1.B. Stimuli used in Chapter 3 ...... 245 1.C. Stimuli used in Chapter 4 ...... 249 1.D. Stimuli used in Chapter 5 ...... 259 1.E. Stimuli used in Chapter 6 ...... 261 1.F. Stimuli used in Chapter 7 ...... 269 Appendix 2 – Questionnaires ...... 285 2.A. Questionnaire used in Chapter 2...... 285 2.B. Questionnaire used in Chapter 3...... 290 2.C. Questionnaires used in Chapter 4 ...... 293 2.D. Questionnaire used in Chapter 5...... 300 v 2.E. Questionnaire used in Chapter 6...... 304 2.F. Questionnaire used in Chapter 7...... 307 Appendix 3 – Alternative analyses ...... 313 3A. Linear regression with actual personalization (Chapter 3) ...... 313 3B. Linear regression with actual personalization (Chapter 4 – Study 1)...... 314 3C. Linear regression with actual personalization (Chapter 4 – Study 2)...... 314 3D. Linear regression with actual message valence, tone of voice and service type (Chapter 6) . 315

Introduction

1 Chapter 1: Introduction

Context and Relevance

People around the globe have embraced social media, and social networking site (SNSs) in particular, as important communication channels: in April 2019 there were 3.48 billion social media users worldwide

(Kemp, 2019). Social media have become important for both their personal and social lives. People primarily use these media for amusement, to pass time or for social exchange (Ku, Chu, & Tseng, 2013).

Advertisers have also found their way to these platforms and try to engage with their (potential) consumers through brand communication. For example, Facebook reported an advertising revenue of more than 50 million dollars in 2018 (Facebook Inc., 2019c). The main objective of this dissertation is to study consumers’ responses to brand communication on social networking sites.

We define brand communication as “any piece of brand-related communication” (Voorveld, 2019, p.

15). This entails communication with and by different stakeholders: for example, the brand itself, the potential, current and past consumers, employees, suppliers, media, government regulators and the community (Duncan & Moriarty, 1998). In that regard, all stakeholders are brand communicators.

Moreover, each consumer is not only a passive information receiver, but can also send brand messages by means of, for example, word-of-mouth or reviews. The media also take a dual role in the brand communication process. On the one hand, they are used to distribute brand communication messages and determine the format of the brand messages (e.g., the size of an advertisement on a social networking site is determined by the site and not by the brand). On the other hand, media themselves are also brands that send out brand communication messages and as such add meaning to the brand communication messages of others. As McLuhan (1964), has put it: “the medium is the message”.

Furthermore, feedback has always been an important part in any communication model (e.g., Lasswell,

1948; Shannon & Weaver, 1949). In traditional marketing theories, communication results in persuasion

(McCarthy, 1960) and feedback, therefore, was obtained through customer service, sales and market research (Duncan & Moriarty, 1998). However, technological developments has increased the possibilities for two-way communication between different stakeholders for example by sending messages on social networking sites.

The current dissertation focuses on two different angles by examining consumer responses to both personalized advertising (firm-initiated brand communication) and word-of-mouth on social networking

2 Chapter 1: Introduction sites (user-generated brand communication). Our studies are set in the context of Facebook, a social networking site which is currently the largest social media site: in June 2019, Facebook had on average 1.59 billion active daily users. We focus primarily on the top and the middle of the (digital) marketing funnel (e.g., Cooler Insights, 2016). When consumers are at the top of the funnel (TOFU) they meet the product or the company, for example by exposure to online ads or storytelling posts (Bátor & Lengyel, 2014). In the middle of the funnel (MOFU) is where awareness become interest and consumers consider buying the product or the brand (Bátor & Lengyel, 2014). In this phase, it is important to expose potential consumers to reviews or to create interesting landing pages (Cooler Insights, 2016). At the bottom of the funnel (BOFU) is where buyers become loyal customers or even advocated of the 1 product or the brand (Bátor & Lengyel, 2014). In this phase, companies can set up contests in which consumers are asked to share brand-related content (e.g. videos or pictures). Moreover, in this phase influencers will share messages about the brand, product or service (Cooler Insights, 2016). To avoid confounds due to prior brand experience our studies are conducted with fictitious brands. These represent situations of a brand launch, where creating awareness and interest are relevant communication goals, as a result most of our research is situated in the top and middle of the marketing funnel.

When online, social networking site users reveal a large amount of personal information, such as their location, their interests, their demographics, etc., creating a digital footprint. Almost all actions (both online and offline) and behaviors, even interactions with friends, are collected by social networking sites such as Facebook (Montgomery, 2015). All that information can be used by advertisers to tailor their messages to specific targets (Kelly, Kerr, & Drennan, 2010; Sundar & Marathe, 2010). For example, Facebook Inc. (2019a) offers the opportunity to specify targets based on their location, demographics, interests, connections, and behaviors. Previous research has also used a number of these personalization elements to elicit perceptions of personalization (e.g., combinations of interests, age, gender and location: Aguirre, Mahr, Grewal, de Ruyter, & Wetzels, 2015; e.g., age and gender: Higgins et al., 2018; e.g., location: Ketelaar et al., 2017). In general, these personalized advertisements tend to be more effective than non-personalized advertisements in terms of visual attention (Bang &

Wojdynski, 2016; Pfiffelmann, Dens, & Soulez, 2019) and in terms of consumers’ attitudes and behavioral intentions (Li & Liu, 2017).

3 Chapter 1: Introduction

You aren’t necessarily aware that when you tell me what music you listen to or what TV shows you

watch, you are telling me some of your deepest and most personal attributes

--Christopher Wylie, Cambridge Analytica whistleblower

These new developments in terms of personalization and the use of personal information have sparked the debate on privacy. Even more so because the year 2018 will be marked as a year full of Facebook scandals. In February 2018, the Mueller indictment revealed how 13 Russian Internet research agents created fake personas on Instagram and Facebook to influence the U.S. election debate (Lapowsky,

2018). On March 18th, 2018 whistleblower Christopher Wylie revealed that Cambridge Analytica, a political consultancy company, used personal information from over 87 million Facebook users to build individual profiles from US voters in order to target them with personalized advertisements (Kang &

Frenkel, 2018). This information was harvested without authorization: Facebook users took a personality test and agreed to have their data collected for academic use. Apart from collecting personal information from Facebook users, the Facebook application also collected information from users’ Facebook connections, accumulating the personal information of millions of Facebook users. That information was not only used for academic purposes but also sold to Cambridge Analytica which targeted US voters based on that information, even though Facebook regulations did not allow for the sales of data obtained from Facebook to third parties (Cadwalladr & Graham-Harrison, 2018). Users strongly reacted to

Facebook and shared the #DeleteFacebook over 10,000 times in just a few hours (Hsu, 2018; Ozcelik

& Varnali, 2019). This scandal resulted in Facebook promising to better protect personal data in the future (Kang & Frenkel, 2018). Nevertheless, since then Facebook had to admit more data breaches

(Lapowsky, 2018). For example, in June 2018 The New York Times reported on deals made between

Facebook and device makers such as Apple, Amazon, Microsoft and Blackberry to share Facebook users’ personal data (Dance, Confessore, & LaForgia, 2018). At more or less the same time as the exposure of a number of Facebook scandals, the European Union announced the new General Data Protection

Regulation, which became effective on May 25th, 2018. This regulation ensures the safekeeping of personal information from European citizens. It entails that companies collecting information from

4 Chapter 1: Introduction citizens need to ask permission to collect and use this information and enforces that companies have procedures for data collection and storage (European Commitee, 2018). This is an important change for many companies, including social networking sites.

It is believed that these scandals and regulations could change consumers’ awareness, and potentially their privacy concerns, which in turn would enable consumers to be more resistant to, for instance, personalized advertising. Nevertheless, Facebook still reports an increase of 8% year-over-year in daily active users in the first quarter of 2019 (Facebook Inc., 2019c), which might indicate that for consumers and advertisers (personalized) social networking site advertising still provides added value. 1 Next to (personalized) advertising, brand communication on social media also encompasses word-of- mouth communication (sWOM), a bidirectional means of brand communication. We define sWOM as electronic word-of-mouth communication (eWOM), “any positive or negative statement made by potential, actual, or former customers about a product or a company, which is made available to a multitude of people and institutions” (Hennig-Thurau, Gwinner, Walsh, & Gremler, 2004, p. 39), on social networking sites. One aspect of modern life is to share daily product/service experiences online

(Kim, Jang, & Adler, 2015). This can occur on review sites (e.g., TripAdvisor), (micro-)blogging platforms

(e.g., Twitter), video sharing sites (e.g., YouTube) and social networking sites (e.g., Facebook). In the current dissertation, we focus on the latter. Social networking sites allow the dissemination of information to a multitude of people and institutions to established social networks through interpersonal online interactions (boyd & Ellison, 2007; Hennig-Thurau et al., 2004). Due to their popularity, consumers often visit them to check for business information and products reviews (Shaw, 2018).

Research on sWOM shows that these messages play an important role in shaping people’s attitudes and behaviors (e.g., Karakaya & Barners, 2010; Rui, Liu, & Whinston, 2013; Schivinski & Dabrowski, 2015;

Wang, Yu, & Wei, 2012).

The main objective of the current dissertation is thus twofold. On the one hand, we will examine how, and under which circumstances, personalized advertising on social networking sites influences consumers attitudes and behavioral intentions. To be able to examine the processing mechanisms of personalized advertising, it is important to know which personalization elements elicit perceived personalization. Therefore, we pose the following research question:

5 Chapter 1: Introduction

RQ1: Which personalization elements elicit perceived personalization?

Boerman, Kruikemeier, and Zuiderveen Borgesius (2017) posit that the processing mechanism of personalized advertising is not yet fully understood. In response to this call, we will look into different mediating variables (i.e., perceived personal relevance, perceived intrusiveness, perceived entertainment, self-brand connection, reactance to the advertisement, perceived creepiness) which can help explain the effects of personalization on consumers’ attitudes and behavioral intentions. Moreover, we will also examine the moderating impact of contextual factors such as attitude toward the social networking site and perceived privacy protection by the social networking site. Therefore, we posit the following research questions:

RQ2: How is (perceived) personalization processed and how does it, subsequently, affect consumers’ attitudes and behavioral intentions?

RQ3: What are boundary conditions of the effects of personalized advertising on consumers’ attitudes and behavioral intentions, and the associated processing mechanisms?

In the second part of the dissertation, we will examine under which circumstances word-of-mouth on social networking sites affect consumers’ attitudes and behaviors. Word-of-mouth on social networking sites can be positive or negative. Most studies on electronic word-of-mouth find that positive eWOM has a positive impact on the recipients’ attitudes (Doh & Hwang, 2009; Purnawirawan, Eisend, De

Pelsmacker, & Dens, 2015), purchase intentions (Bae & Lee, 2011; Doh & Hwang, 2009) and sales

(Floyd, Freling, Alhoqail, Cho, & Freling, 2014), while negative eWOM results in negative effects. More specifically with regard to sWOM (electronic word-of-mouth on social networking sites), Rui et al. (2013) report a positive effect of positive tweets about a movie on movie sales and a negative effect of negative tweets. This message valence effect appears to be consistent. It can, however, be strengthened or weakened by moderating factors. The current dissertation aims to contribute to the understanding of the circumstances under which this message valence effect affects consumers’ attitudes and behaviors.

We study the moderating impact of the tone of voice (factual or emotional) of the word-of-mouth message on social networking (sWOM) and the type of service (hedonic or utilitarian) mentioned in the sWOM message. Furthermore, we study the moderating impact of interpersonal and person-to-site

6 Chapter 1: Introduction relational characteristics on how message valence affects consumer responses. Therefore, our fourth, and final, research question is as follows:

RQ4: What are boundary conditions of the message valence effect on consumers’ attitudes and behavioral intentions?

The following sections of this chapter provide an overview of the content and the structure of the dissertation and the theoretical frameworks used throughout the dissertation.

Structure of the dissertation The dissertation consists of six empirical chapters, which are graphically presented in Figure 1.1. 1 Chapters 2 – 5 examine personalized advertising. In these chapters, we gradually build up an understanding of how personalized advertising is being processed and how it affects consumers attitudes and behavioral intentions. One of these behavioral intentions is word-of-mouth. The last two chapters focus on word-of-mouth on social-networking sites.

The first objective of this dissertation is to determine the relative importance of different personalization elements to elicit perceived personalization. Chapter 2 explores how to elicit perceived personalization. As mentioned in the previous section, it is unclear when the manipulation of personalization elements leads to a perception of personalization. However, it is this perception which drives the effects of personalization (Kramer, 2007; Li, 2016). Previous research uses different personalization elements (e.g., demographics or location) to manipulate personalization (Higgins et al.,

2018; Ketelaar et al., 2017). However, not all studies report a manipulation check. Others use combinations of personalization elements making it impossible to determine how much each element contributes to the perception of personalization elements (Aguirre et al., 2015; Zarouali, Poels, Walrave,

& Ponnet, 2018). Furthermore, different measures are used to measure perceived personalization making it difficult to compare. Therefore, this chapter investigates the relative importance of six personalization elements (age, gender, location, life events, interests, and page-likes of friends) in eliciting perceived personalization and tests the robustness of this relative importance over products and audience characteristics. Participants in a conjoint analysis rated the perceived personalization of six advertisements for one of six products or services.

7 Chapter 1: Introduction

The second objective of this dissertation is to explain the positive and negative effects of personalized advertising on consumers’ attitudes and behavioral intentions. We gradually disentangle the underlying processing mechanisms of personalized advertising on consumers’ attitudes and behavioral intentions. These underlying processes are not well established (Boerman et al., 2017).

Chapter 3 examines the mediating effect of personal relevance. This is documented in previous research in other contexts and is, according to Noar, Harrington, and Aldrich (2009), the most important mediator in the relationship between personalized messages and consumers’ attitudes and behavioral intentions.

Chapter 3 examines the effect of perceived personalization on brand attitude and click-intention through perceived personal relevance and tests the moderating impact of the attitude toward the social networking site. Participants in two experiments are exposed to a personalized advertisement (based on gender) which featured a fictitious advertisement for perfume (Study 1) and a vacuum cleaner (Study

2). Both studies are based on a between-subjects design with two experimental conditions (personalized vs non-personalized).

In Chapter 4, the potentially negative explanatory mechanism via perceived intrusiveness is explored.

This mediator has been reported by van Doorn and Hoekstra (2013) and by White, Zahay, Thorbjørnsen, and Shavitt (2008) to be of crucial importance in explaining the negative effects of personalization on consumer responses. We study the role of intrusiveness together with personal relevance to assess their relative importance. More specifically, Study 1 of Chapter 4 examines the effects of perceived personalization on brand attitude and click intention. This first study employed a between-subjects experiment in which participants are exposed to one of four possible advertisements. The advertisement could be non-personalized or personalized based on either gender, age or interests. In Study 2 of

Chapter 4, more mediating variables are added to improve the understanding of the effects of personalization of brand attitude, click intention and word-of-mouth intention. The mediator ‘perceived entertainment’ is added to the model, which refers to consumers’ perception of the extent to which an advertisement provides pleasure, diversion or amusement (Taylor, Strutton, & Thompson, 2012).

Together with personal relevance and perceived intrusiveness, perceived entertainment is considered as a first-order mediator. We also add two second-order mediators to the model: self-brand connection and reactance to the advertisement. Self-brand connection refers to the connection consumers establish between their self-concept and the brand (Liu & Mattila, 2017) and, thus, might be able to account for

8 Chapter 1: Introduction the positive effects of personalized advertisements. Reactance to the advertisement would then account for the negative effects and refers to a psychological reactance consumers might have when encountering an advertisement that is too personal (White et al., 2008). These mediating variables are examined in a second between-subjects experiment in Chapter 4, in which respondents are exposed to one of five advertisements. To create the four personalized advertisements – the fifth advertisement was not personalized - we construct combinations of the personalization elements gender and age, interests, life events and page-likes.

We explore potentially negative personalization effects further in Chapter 5 by investigating the effect of perceived creepiness, next to personal relevance, on the attitude toward the source of the 1 advertisement. Perceived creepiness has several definitions (see e.g., Barnard, 2014; Malheiros,

Jennett, Patel, Brostoff, & Sasse, 2012; McAndrew & Koehnke, 2016; Zhang & Xu, 2016). In our case, the relevant creepiness factor is the fact that personalized advertising could result in feelings of uncertainty and unpredictability. In a 3 (level of personalization: low, moderate, high) x 4 (source: health website, governmental website, online newspaper and commercial website) between-subjects experiment respondents are exposed to a vignette about an advertisement on a social networking site.

The third objective of this dissertation is to examine boundary conditions of the effects of personalized advertising on consumers’ attitudes and behavioral intentions and the associated processing mechanisms. The dissertation first focuses on context-related potential moderators: the attitude toward the social networking site and perceived privacy protection by the social networking site. The attitude toward the website has been found to affect consumers’ responses to advertising embedded in the website (Stevenson, Bruner, & Kumar, 2000). Research in SNS also shows that a favorable attitude toward a social networking site can reinforce positive consumers responses toward advertised products (Wen, Tan, & Chang, 2009). Chapters 3 and 4 study the moderating role of the attitude toward the social networking site. In Chapter 4, the role of perceived privacy protection by the social networking site is also explored. Previous research indicates that when consumers do not feel informed about a company’s privacy policies, they will be more reluctant towards providing personal information (Malhotra, Kim, & Agarwal, 2004). Moreover, users who trust a website, because it provides information on privacy protection, exhibit higher click intention when encountering a personalized

9 Chapter 1: Introduction advertisement on that website compared to when the advertisement is embedded in an untrustworthy website (Aguirre et al., 2015). Nevertheless, previous research did not examine the moderating role of these variables on both the direct and indirect effects of personalized advertising in social networking sites on consumer responses.

In Chapter 5, the impact of the type of source (i.e., the advertiser) is also examined. Studies examining the effects of personalized advertising from different sources are rare (Bol et al., 2018; Smit, van Noort,

& Voorveld, 2014; Ur, Leon, Cranor, Shay, & Wang, 2012). In line with previous literature, we expect that for some advertisers (e.g., news website), the use of personalized advertising is more acceptable than for others (e.g., health website), because consumers have learned that the use of personal information is not always appropriate (Acquisti, Brandimarte, & Loewenstein, 2015).

The fourth objective of this dissertation is to examine boundary conditions of the message valence effect of sWOM messages on consumers’ attitudes and behavioral intentions. The general message valence effect is well-established in previous research (Purnawirawan et al., 2015). In

Chapter 6, we look into message content factors of WOM messages on social networking sites that could moderate this effect. Barger, Peltier, and Schultz (2016) suggest that these factors need more extensive research. This dissertation responds to that call by inspecting how the tone of voice of a message

(whether the sWOM message takes on a more factual or emotional tone) affects the effect message valence has on the attitude toward the service provider, purchase intention and positive WOM intention. sWOM messages can be (predominantly) factual, based on attribute-value information. The arguments used are rational, objective, specific and clear. Other messages are (predominantly) emotional, focusing on the feelings that result from the experience of using the product or service, lacking verifiable arguments (Park & Lee, 2008). The accessibility-diagnosticity theory suggests that factual messages, which are based on concrete product characteristics, are perceived to be more informative and diagnostic. Also, the attribution theory (Moran & Muzellec, 2014; Sen & Lerman, 2007) suggests that factual messages will stimulate stimulus attribution – attribution based on the product performance - which results in a better insight in product performance. Emotional messages, on the other hand, are more likely to stimulate non-stimulus attribution: a lack of arguments and emotional statements will result in attributing the product performance to the dispositional characteristics of the communicator.

10 Chapter 1: Introduction

In Chapter 6, we also study the moderating effect of service type (whether the service primarily fulfills utilitarian or hedonic buying motivations) on the effect of sWOM. The accessibility-diagnostic theory suggests that information about a hedonic product or service will be more diagnostic than for a utilitarian product or service (Willemsen, Neijens, Bronner, & de Ridder, 2011). Hedonic products and services

(e.g. restaurant, bar) are primarily used for the affective and sensory experience they provide (Hirschman & Holbrook, 1982). They are highly person-specific and therefore mostly experienced subjectively (Voss, Spangenberg, & Grohmann, 2003). Due to the fact that they cannot be evaluated before experience or usage (Klein, 1998; Zhu & Zhang, 2010), the pre-purchase risk is high (Park & Lee, 2009; Park & Park, 2013), which might explain the higher diagnosticity of sWOM for hedonic 1 products. Utilitarian products and services (e.g., cell phone provider, bank), on the other hand, are primarily used to accomplish a functional or practical task. As such, they can be evaluated based on tangible attributes prior to purchase (Klein, 1998; Purnawirawan et al., 2015). They are less person- specific and, therefore, their evaluation can be made more objectively (Voss et al., 2003) and sWOM messages will be relatively less diagnostic for utilitarian products (Willemsen et al., 2011). Additionally, based on the matching principle (e.g., Klein & Melnyk, 2016), an effect of the match between the message tone of voice and service type is expected. In Chapter 6 respondents are exposed to an sWOM message in a 2 (message valence: positive vs negative) x 2 (tone of voice: factual or emotional) x 2

(service type: utilitarian vs hedonic) between-subjects experiment. sWOM does not occur in a social vacuum: the relationship between the receiver and the sender of the message is part of the context in which the interaction takes place (Chu & Kim, 2011; Kim, Park, Lee, & Park, 2018). Therefore, it is crucial to understand how relational characteristics, such as homophily (similarity between sender and receiver), tie strength (strength of the relationship between sender and receiver) and source credibility (evaluation of the sender’s credibility by the receiver) affect the effect of message valence on consumer responses. Brown, Broderick, and Lee (2007) argue that, in online social networks, individuals develop relationships with the site on which they are communicating rather than with other individuals using the site. This proposition comes across as somewhat counterintuitive and has not been empirically validated. Kim, Kandampully, and Bilgihan (2018) offered a partial test of the proposition by studying the effects of homophily, tie strength and source-credibility on the ability of sWOM to influence purchase decisions (sWOM effectiveness) and show that the person-to-site variables

11 Chapter 1: Introduction indeed explain some of the variance in sWOM effectiveness. However, this study does not take the interpersonal relational characteristics into account. Chapter 7 studies the person-to-site relational characteristics in the presence of interpersonal relational characteristics. In this chapter respondents are exposed to a 2 (message valence: positive vs negative) x 2 (interpersonal homophily: high vs low) x 2 (interpersonal tie strength: high vs low) x 2 (interpersonal source credibility: high vs low) between- subjects experiment. Person-to-site homophily, tie strength, and source credibility were not manipulated but measured, because Facebook is a site with which most participants are highly familiar and for which people naturally vary in their perceptions of the three variables under study.

Finally, Chapter 8 presents a summary of the main findings and discusses the implications for theory and practice. A number of limitations and suggestions for future research are presented.

It must be noted that all chapters are written as independent stand-alone articles (either already published, currently under review, or ready for submission to academic journals). As a result, there is some overlap between the current introductory chapter, the literature review of each chapter and the conclusion of this dissertation

Figure 1.1. Overview of dissertation

Chapter 1: Introduction Chapter

Chapter 1: Introduction

Theoretical background

Part I: Personalized advertising in social networking sites

Definition

The concept of personalized advertising is not new or limited to social networking sites. Researchers use different terms (e.g., targeting, tailoring, personalization, customization) and different definitions to describe it (Kim & Sundar, 2012; Vesanen, 2007). Targeting implies that specific segments in the market are targeted with advertising messages (Iyer, Soberman, & Villas-Boas, 2005). Tailoring involves adapting persuasive messages to recipients characteristics (Hirsch, Kang, & Bodenhausen, 2012). There are two types of tailoring: personalization and customization (Arora et al., 2008; Lee, Kim, & Sundar,

2015). Customization involves user-initiated information tailoring, which means that users themselves specify the type of content they want to receive. Users can use personal settings to inform the system of their needs and interests (Lee et al., 2015). Personalization, on the other hand, involves system- initiated information tailoring, which means that a system tracks a user’s demographics (e.g. age, gender), preferences (e.g. preferred product and services) and geographic information and uses that information to tailor the advertising message (Lee et al., 2015). The underlying idea is, thus, that advertising is delivered tailored to an individual’s characteristics and/or interests or tastes (Hoy & Milne,

2010; Kelly et al., 2010). As such, personalization can range from no personalization at all, when the consumer receives a generic message, over general personalization, when broad categories (e.g., sending baby clothing ads to women who recently indicated on Facebook they had a baby) are used to personalize the message, to full personalization, when the advertising message is addressing one specific individual based on, for example, his name or connections to specific pages, groups or applications (Arora et al., 2008; Hawkins, Kreuter, Resnicow, Fishbein, & Dijkstra, 2008; Hoy & Milne,

2010).

The effects of personalized advertising on attitudes and behavior(al) intentions

Previous research shows that (perceived) personalization can improve advertising effectiveness (e.g. ad attitude, brand attitude, patronage intention: Maslowska, Smit, & van den Putte, 2016; e.g., more and longer visual attention: Pfiffelmann et al., 2019). These positive effects are frequently attributed to an increase in personal relevance (Kim & Huh, 2017; Sundar & Marathe, 2010; Zhu & Chang, 2016). On the other hand, research also suggests negative effects of personalized advertising, which are attributed

14 Chapter 1: Introduction to intrusiveness (van Doorn & Hoekstra, 2013; White et al., 2008) or creepiness (Malheiros et al., 2012;

Ur et al., 2012) of the advertisement. As becomes clear from these mixed findings, the underlying processes (both beneficial and detrimental) explaining personalized advertising effectiveness, remain an area for further research (Boerman et al., 2017).

A number of scholars indicate that in order for personalization to affect consumers’ responses, consumers need to perceive the advertisement as personalized (Kramer, 2007; Li, 2016). Actual personalization, which refers to how personalization is manipulated by advertisers or researchers, does not always match with perceived personalization, which refers to the perception an individual has about the match between the advertisement and his/her own self-image (Li, 2016). Consequently, perceived 1 personalization is a more relevant construct than actual personalization (Kramer, 2007). Moreover, by using consumers’ perceptions we are less dependent of the specific experimental conditions which potentially affect our conclusions and, therefore, we are able to draw more valid conclusions across the different experiments (and experimental conditions). Therefore, we have used consumers’ perceptions in all but one chapter.

Privacy calculus and personalization paradox

The notion of privacy calculus can be used to describe the process in which consumers weigh the benefits and the costs (or risks) of personalized advertising. Personalization techniques have become more data-driven and the collection and processing of these data are a pivotal element in the process of personalizing advertising. In turn, consumers are increasingly aware of the costs and risks of their privacy (Aguirre et al., 2015; Bol et al., 2018). The potential benefit (e.g., personal relevance, entertainment) might be outweighed by the risks and costs (e.g., intrusiveness, creepiness) and consumers might respond negatively to it. The privacy calculus theory has often been applied to explain self-disclosure behavior (Culnan & Armstrong, 1999; Dienlin & Metzger, 2016; Laufer & Wolfe, 1977).

However, it might also affect how consumers respond to advertisers using that self-disclosed information

(Demmers, van Dolen, & Weltevreden, 2018). This calculus can also be seen in the light of the personalization paradox (Awad & Krishnan, 2006): personalization increases the relevance and usefulness of the advertising message. On the other hand, consumers also become more vulnerable in terms of privacy (Aguirre et al., 2015; Bol et al., 2018). Previous research suggests that, even though

15 Chapter 1: Introduction consumers express privacy concerns, they still disclose a lot of personal information (Aguirre et al.,

2015).

Part II: Word-of-mouth on social networking sites

Definition

Word-of-mouth (WOM) is not a new phenomenon or limited to social networking sites. Katz and

Lazarsfeld (1955) already indicated that it was seven times more effective than advertising to persuade consumers to switch brands. It is believed to convey unbiased information about a product or a service which results from the fact that the sender is non-commercial (Willemsen, 2013). Therefore, WOM is considered to be more credible and useful, and ultimately, more persuasive than marketer-initiated communication (Sweeney, Soutar, & Mazzarol, 2008). Traditional WOM has its constraints, both in terms of time and place: the amount of people with which a WOM message can be shared is limited due to the fact that the receiver of the message physically meets the sender of the message. Moreover, the message is sent at a specific moment in time and the receiver is not able to retrieve it at a later time.

These constraints have become less important with the advent of the internet and even more so with the rise of social networking sites. It is important to distinguish word-of-mouth on social networking sites (sWOM) from electronic word-of-mouth (eWOM). Previous research has made it clear that not all digitally distributed word-of-mouth results in the same outcomes (e.g. Lin, Hsu, Chen, & Fang, 2017).

On the one hand, sWOM senders are identifiable and can be held accountable, therefore, the social risk of sWOM is much larger compared to eWOM (Balaji, Khong, & Chong, 2016; Eisingerich, Chun, Liu, Jia,

& Bell, 2015). On the other hand, this also ensures that a reader can assess the credibility (from the source’s personal SNS profile), which in turn positively affects the likelihood that the information in the sWOM message will be used by consumers in their evaluation of the product or service in the message

(Hajli, 2016).

Accessibility-diagnosticity model

The accessibility-diagnosticity model (Feldman & Lynch, 1988) is often used to explain the effects of

WOM, and thus also sWOM, messages (Herr, Kardes, & Kim, 1991). In order for consumers to use any piece of information, the information must be available in consumers’ memory (accessible) as well as be diagnostic. Accessibility refers to the ease of retrieving information from memory (Menon, Raghubit,

16 Chapter 1: Introduction

& Schwarz, 1995). A piece of information is perceived as diagnostic if it enables the consumer to assess the quality and the performance of a target object (e.g., the product reviewed in an sWOM message). Whether or not information from an sWOM message is used to asses a product or service, depends on (a) whether the information is accessible, (b) whether the information is diagnostic and (c) whether there is other accessible, more diagnostic information present (Herr et al., 1991; Menon et al., 1995). According to the model, when more diagnostic information is available not all accessible information is used as a basis for evaluation and choice (Feldman & Lynch, 1988). Diagnostic information that is less ambiguous will more likely be used as it enables a more straightforward evaluation of the product or service at hand. Mudambi and Schuff (2010) posit that information with high perceived diagnosticity 1 enables consumers to make better purchase decisions, which indicates that the relevance of the available information is what helps consumers evaluate a product or service.

Attribution theory The attribution theory (Kelley, 1973) has also been used to explain the effects of WOM (Laczniak,

DeCarlo, & Ramaswami, 2001; Moran & Muzellec, 2014; Sen & Lerman, 2007). “According to this theory, people have the tendency to make sense of their surroundings by acting as naive psychologists” (Eberly, Holley, Johnson, & Mitchell, 2011, p. 731). When they are confronted with an sWOM message, they try to determine the reason for the information in the message, which is the senders’ evaluation of the product or the service in the message. For example, when the sender is negative about the hotel he just stayed in, the receiver will try to determine what caused the negative experience. These causal attributions, made in response to information a consumer receives, can be based on the stimulus (i.e., the brand or the product or service), the person (i.e., the sWOM sender), the circumstance (e.g., the sWOM message reports that the elevators were broke which was due to a power shortage in the area) or a combination of these three. When the reported product performance in an sWOM message is attributed to the stimulus, it will be more important for consumers’ evaluations of the brand, product or service than compared to when it is attributed to, for example, the sWOM sender (Laczniak et al., 2001).

Relational characteristics

Previous research examined the effects of offline word-of-mouth on consumers attitudes and behavioral intentions in the light of the social network in which the WOM exchange took place (Bansal & Voyer,

17 Chapter 1: Introduction

2000; Brown & Reingen, 1987). WOM exchange is always embedded in a social relationship between at least two actors. According to Brown and Reingen (1987), the ties, or the link, between communicators are crucial to understand the process of WOM.

A first, often mentioned, characteristic is interpersonal homophily, which refers to the extent to which pairs of individuals are alike and share, for example, age group, gender, lifestyle, social style, or interests

(Brown et al., 2007; Rogers & Bhowmik, 1970). Even though not all of these cues are present online, previous research found that perceptions of interests and opinions are used to evaluate online interpersonal homophily (Blanton, 2001; Gilly, Graham, Wolfinbarger, & Yale, 1998). Moreover,

Festinger (1954) suggests that people implicitly assume that individuals similar to themselves have similar needs and preferences, and thus they tend to compare themselves with others who are similar.

As a result, interpersonal homophily stimulates the likelihood that information in WOM messages will be used.

Another characteristic of the relationship between communicators is tie strength, which refers to the closeness of the relationship between two actors (Brown et al., 2007). It ranges from weak to strong ties and depends on the importance attached to the relationship, the frequency of contact, the type of social relation and the intimacy between the communicators (Brown et al., 2007; Granovetter, 1973;

Marsden & Campbell, 1984). WOM sent by a strong tie, compared to a weak tie, has been found to have a greater impact on the receivers’ behavior (Brown et al., 2007; De Bruyn & Lilien, 2008), because the information is perceived a more trustworthy, and thus reduces potential risks (De Bruyn & Lilien,

2008; Shan & King, 2015).

The third characteristic that is often used to understand WOM effects, is source credibility. It consists out of two dimensions: expertise – the perceived competence of someone within a certain domain - and trustworthiness – the degree to which someone can be trusted and to, for example, not act out of self- interest (Ohanian, 1990). Credible sources are more persuasive (Teng et al., 2014) because they increase the likelihood that attention is being paid to the message, that the message is perceived as useful (Li, Huang, Tan, & Wei, 2013), reliable (Cheung, Lee, & Rabjohn, 2008) and credible (Cheung,

Luo, Sia, & Chen, 2009).

PART I - Personalized advertising -

Instead of a one-way interruption, personalized marketing is about delivering value at just the right

moment that a user needs it.

--David Meerman Scott, marketing strategist

Let’s Get Personal: Which Elements Elicit Perceived Personalization in Advertising1

1 Earlier versions of this chapter were presented at 2018 Interactive Marketing Research Conference in Amsterdam, The Netherlands and at 2019 International Conference on Research in Advertising, Krems an der Donau, Austria 2 Chapter 2: Let’s get personal

Abstract

On social networking sites, consumers disclose information about themselves that advertisers use to personalize advertisements. The underlying assumption is that personalized advertisements are more persuasive. However, it is not clear to what extent actual personalization elements (as intended by advertisers) determine consumers’ perceptions of the extent to which an ad is personalized, and it is the latter that drives responses. The current study investigates the relative importance of different actual personalization elements (age, gender, location, life events, interests, and friend referrals) in Facebook ads in eliciting perceived advertising personalization. We conduct conjoint analyses for six products (a bank, a smartphone, tableware, furniture, a restaurant, a fashion retailer) with 595 U.S. consumers.

The findings show that the most important elements in eliciting perceived personalization are (in order of importance) a person’s interests, location, and age. This result remains stable across different product perceptions (product category involvement, product qualities and buying motivations) and across different socio-demographic characteristics (gender, age, and education).

22 Chapter 2: Let’s get personal

Introduction

In April 2019, there were 3.48 billion social media users, which was an increase of 9% compared to

2018 (Kemp, 2019). Facebook, the world’s most popular social networking site, now has 2.41 billion monthly active users (Facebook Inc., 2019b, 2019d). This increasing use of social networking sites has also caught advertisers’ attention (De Keyzer, Dens, & De Pelsmacker, 2015). SNSs are increasingly considered as an interesting platform for firms to reach their target groups because it is faster than traditional media (Saxena & Khanna, 2013; Trusov, Bucklin, & Pauwels, 2009) and specific target groups can be efficiently reached (Kelly et al., 2010; Sundar & Marathe, 2010). For instance, Facebook Inc. (2019c) reports that its advertising revenue rose from almost $27 billion in 2016 to $55 billion in 2018.

People disclose a lot of personal information, such as their location, interests, and demographic 2 characteristics, on SNSs. Advertisers can use this information to tailor their messages to specific targets

(De Keyzer et al., 2015; Kelly et al., 2010; Sundar & Marathe, 2010). Advertising that is tailored to one’s characteristics and/or interests is called personalized advertising (Hoy & Milne, 2010; Kelly et al., 2010;

Sundar & Marathe, 2010). Personalized advertising tends to be more persuasive than non-personalized advertising: it generates longer and more visual attention (Bang & Wojdynski, 2016; Pfiffelmann et al.,

2019) and tends to result in a more positive brand attitude and higher click intentions because consumers perceive it as more relevant (De Keyzer et al., 2015). Moreover, personalized advertising is processed more carefully and will more likely be used for decision-making (Rimer & Kreuter, 2006).

Previous studies have indicated that manipulated or actual personalization (as intended by the advertiser based on previous data) is not always perceived by the consumer as being personalized (De Keyzer et al., 2015; Kramer, Spolter-Weisfeld, & Thakkar, 2007; Simonson, 2005). Perceived personalization is dependent on the perception of the extent to which the message receiver perceives the message to fit his personal profile (Li, 2016). Kramer (2007) posits that consumers should perceive a message as tailored to their needs and preferences before any effect of personalization can take place. Little is known, however, about the extent to which different actual personalization elements contribute to consumers’ perceptions of personalization. Prior research has manipulated personalization in different ways, based on a variety of elements such as gender (De Keyzer et al., 2015), location (Ketelaar et al.,

2017) or various combinations (e.g., interests, age, gender and location, Aguirre et al., 2015). Many

23 Chapter 2: Let’s get personal studies fail to report whether their manipulations successfully resulted in a greater degree of perceived personalization and when they do report a manipulation check, different studies use different measures.

In studies that combine elements, it is also impossible to disentangle the effect of individual personalization cues. The first purpose of the present study is to explore the relative importance of six frequently used personalization elements (location, age, gender, life events, interests and friend referrals) in triggering perceived personalization in consumers.

Additionally, the relative importance of actual personalization elements for perceived personalization may differ across perceived product characteristics (particularly, product category involvement, search vs. experience qualities, or hedonic vs. utilitarian buying motivation), as differences in motivations for purchasing products may lead people to attach importance to distinct advertising elements.

Furthermore, the relative importance of personalization elements may also differ across different demographic segments (such as men vs women, younger vs older, lower vs higher education), due to differences in processing. Figure 2.1. presents our conceptual framework.

Figure 2.1. Conceptual framework

To investigate these research questions, we conducted a conjoint experiment with U.S. consumers (n

= 595), in which the relative importance of six personalization elements (location, age, gender, life events, interests, and friend referrals) for perceived personalization is explored across six different products and services (a fashion retailer, a restaurant, a smartphone, a bank, tableware, and furniture).

The study adds to the understanding of how personalization elements affect personalization perceptions and the role of product characteristics and demographic characteristics in this process. Advertising

24 Chapter 2: Let’s get personal

practitioners and researchers can use the insights of the study to tailor their personalized advertisements more effectively.

Theoretical Background Different Types of Personalization Elements To determine the different types of personalization elements we use four types of personalization elements that are defined by Facebook Inc. (2019a): location, demographics, interests and friend referrals. We do not include behavior, which is a fifth type mentioned by Facebook, due to the fact that “the tracking of online activities, collection of behavioral data, and dissemination of information often happens covertly” (Boerman et al., 2017, p. 364). Moreover, it often happens outside of the Facebook environment, making it very different from the other types. The following section gives an overview of 2 previous literature that has used one (or more) of the four personalization elements that we selected.

Each of the elements is expected to increase perceived personalization. However, it is not clear how strongly each element is likely to contribute. Different researchers have manipulated personalization for different products and using different elements. Moreover, comparing these studies is difficult because different studies have used different measures for perceived personalization. Table 2.1. provides an overview of experimental studies using different actual personalization elements to manipulate personalization and different measures for perceived personalization.

26 Table 2.1. Overview of experimental studies manipulating personalization personal get Let’s 2: Chapter

Authors (year) Manipulation of personalization Manipulation check? Measurement of perceived personalization Kalyanaraman and Sundar 1. No personalization (0% matched to respondent’s Yes, successful.  The content and information featured on (2006) reported interests (e.g., sports, movies and books)) the website targeted me as a unique 2. 41.7% of the customizable elements matched to individual. interests  This website was personalized according 3. 100% of the customizable elements matched to to my interests interests Sheng, Nah, and Siau (2008) 1. No personalization Yes, successful  "In this scenario, Company X offers 2. Combination of location and preferences personalized services based on my location.  In this scenario, Company X offers personalized services based on my preferences.

 In this scenario, to what extent has Company X personalized its weather service to you?" Sundar and Marathe (2010) 1. Interests (e.g., environmental issues) Not reported. Not reported. 2. Customized webpage (could be rearranged by respondent) 3. Standard webpage Malheiros et al. (2012) 1. General ads Not reported. Not reported. 2. Ads based on preferred holiday destination 3. Ads based on age or name or participants’ photo van Doorn and Hoekstra 1. Browsing data Not reported. Not reported. (2013) 2. Combination of browsing data and name 3. Combination of browsing data and information about previous actions Tucker (2014) 1. No personalization Not reported. Not reported. 2. Combination of education and interests Aguirre et al. (2015) 1. No personalization Yes, successful.  This advertisement is directed to me 2. Interests (based on a scenario provided by the personally. researchers)  I recognize my personal situation in this 3. Interests from condition 2 combined with age, advertisement. gender and location  This advertisement takes into account the problem I faced.  This advertisement takes into account my personal situation. Bleier and Eisenbeiss (2015) 1. Products from most viewed product category Not reported. Not reported. 2. Products from most viewed brand

3. Combination of most viewed product category and brand Bodoff and Ho (2015) 1. No personalization Yes, successful. Not reported. 2. Education: study major, self-reported knowledge level and study stage De Keyzer et al. (2015) 1. No personalization Not reported Not reported. 2. Gender Brinson and Eastin (2016) 1. No personalization Not reported. Not reported. 2. Combination of location and gender Bang and Wojdynski (2016) 1. No personalization Not reported. Not reported. 2. Combination of name and location Kim and Gambino (2016) 1. No personalization Yes, successful.  I feel the content on Foodies.com was 2. Combination of interests (type of cuisine) and tailored to my preferences of foods. label “Foodies.com choice for YOU” Li (2016) 1. No personalization Yes, successful.  The ad seems to be designed specifically 2. Interests for me.  The ad targeted me as a unique individual. Maslowska et al. (2016) 1. Raising expectations: “This message is just for Only identification and combining  Did you have a feeling that you were you.” the types significantly increase addressed personally in the newsletter? 2. Identification: mentions the respondent’s name personalization.  Was the newsletter targeted at you? three times  Could you recognize yourself in the 3. Contextualization: based on gender and group the newsletter was targeted at? occupation (student) 4. Combination of the 3 elements Song, Kim, Kim, Won Lee, and 1. Name of a fictitious consumer Yes, successful.  This e-mail message was tailored to my Lee (2016) 2. Scenario-based information from Tweet, age, and need. occupation.  This e-mail message gives me personal perso get Let’s 2: Chapter attention.  This e-mail message knows what I want. Walrave, Poels, Antheunis, 1. First name only Not reported. Not reported. Van den Broeck, and van 2. Combination of first name, gender, birthday, Noort (2016) interests and profile picture 3. Combination of first name, gender, birthday, interests, surname, address, email address, album picture, and status update Ketelaar et al. (2017) Location of advertised product in store Not reported. Not reported.

1. was incongruent with consumers’ location in the nal

27 store

2. was congruent with consumer’s location in the store

28 Li and Liu (2017) 1. No personalization Yes, successful.  The ad seems to be designed specifically personal get Let’s 2: Chapter

2. Greeting by name for me.  The ad targeted me as a unique individual. Matz, Kosinski, Nave, and Personality traits (study 1: extraversion; study 2: Not reported. Not reported. Stillwell (2017) openness) 1. No match: ad personality was extraverted/open and audience personality was introverted / not open 2. No match: ad personality was introverted / not open and audience personality was extraverted/open 3. Match: ad personality was extraverted/open and audience personality was extraverted/open 4. Match: ad personality was introverted / not open and audience personality was introverted / not open

Tran (2017) Scenario-based retargeting (for either a laptop or Not reported. Not reported. shoes) De Keyzer, Dens, and De 1. No personalization Not reported.  This advertisement makes Pelsmacker (2018) 2. Gender recommendations that match my needs. 3. Age  This ad informs me about products that 4. Interests are tailor-made for me.  This ad is tailored to my situation.  I believe that this ad is customized to my needs.  This advertisement was personalized according to my interests.  This advertisement was targeted at me as a unique individual. Higgins et al. (2018) 1. Age Not reported. Not reported. 2. Gender 3. A combination of age and gender Karmakar and Webster (2018) 1. No personalization Yes, successful.  The presented environmental material: 2. Interests  matched my personal interest;  was tailored to my interest;  matched my personal concern;  was personalized to my personal preference;  featured information related to my specific interests;  personalized environmental degradation for me;

catered to my personal interests; spoke to my personal interest. Zarouali et al. (2018) 1. Interests Yes, successful.  Was the advertisement personalized 2. combination of interests, age, gender and according to your personal interests? location Windels et al. (2018) 1. No personalization Not reported. Not reported. 2. Friend referrals De Keyzer, Kruikemeier, and 1. Friends’ page-likes Not reported. Not reported. van Noort (2019) 2. Combination of friends’ page-likes and gender 3. Combination of friends’ page-likes, gender and interests. Kaspar, Weber, and Wilbers 1. Non-personalized: opposite of respondent’s Not reported. Not reported. (2019) characteristics 2. Personalized: combination of gender, age, professional interest, place of residence and current occupation

Chapter 2: Let’s get perso get Let’s 2: Chapter

nal

Chapter 2: Let’s get personal

Location

The market can be divided based on geographical units, such as nations, states, and cities (Kotler &

Keller, 2006). As can be seen from Figure 2.2, these geographical areas can be used to select specific target groups: it shows an advertisement from Coca-Cola bottling Co. Consolidated which was directed to Facebook users in Indiana and Kentucky explicitly referring to the city of Louisville in the advertisement.

Figure 2.2. Facebook advertisement using location by Coca-Cola Bottling Co. Consolidated

Note. Reprinted from Facebook Ad Library, 2019. Retrieved from https://www.facebook.com/ads/library/?country=US&view_all_page_id=92687984104

Previous research has used consumers’ geographical location to manipulate personalization. In a study on location-based advertising, Ketelaar et al. (2017) do not report whether location-congruent ads (ads appearing when the consumer is in the same location as the advertised product in a virtual supermarket) are perceived as more personalized than location-incongruent ads (when the consumer is not in proximity of the advertised product). While they find that the former leads to a higher brand choice

(which could indicate that location contributes to perceptions of personalization), there is no significant effect of location congruence on ad attention (which would indicate the contrary).

30 Chapter 2: Let’s get personal

Other studies using location to manipulate personalization have used location concurrently with other personalization elements. Some of these studies do not report a manipulation check of perceived personalization (Bang & Wojdynski, 2016; Brinson & Eastin, 2016; Kaspar et al., 2019; Walrave et al., 2016). For example, Bang and Wojdynski (2016) include the respondent’s name and the nearest location of the advertised coffee shop in their advertisement (by showing the coffee shop on a map in the advertisement). Their eye-tracking study shows that a personalized advertisement receives significantly more attention (total amount of time of fixations) than a non-personalized advertisement as well as increases the number of visits to the advertisement. The increase in attention would suggest that consumers indeed perceive the ad as more personalized. Using a scenario-based method, Sheng et al. (2008) manipulate personalization of an online weather service by not only providing real-time weather reporting based on the customer’s location but also by mentioning that they can set their preferences 2 for being alerted to severe weather conditions (e.g., tornados, hurricanes). Zarouali et al. (2018) distinguish between ‘low’ and ‘high’ personalization by adding information about respondents’ age, gender, and location to adolescents’ general interest for sneakers. While the two latter studies both report that their manipulations were successful in increasing the perceived personalization, it is not possible to disentangle the effects of location as a personalization cue from those of the other cues they used.

Demographics Demographic characteristics such as gender, age, or people’s stage in the life cycle often form a base for market segmentation (Kotler & Keller, 2006). They are therefore also often used to personalize advertisements. Figure 2.3 shows an advertisement from High on LifeHigh on Life (2019), a personal training programme, looking for women between the age of 45 and 55 from Edegem, a Belgian city to participate. This advertisement explicitly refers to age and gender (as well as location) as personalization elements.

31 Chapter 2: Let’s get personal

Figure 2.3. Facebook advertisement using age, gender and location from High On Life

Note. Reprinted from Facebook Ad Library, 2019. Retrieved from https://www.facebook.com/ads/library/? country=BE& view_all_page_id=340411786304159

De Keyzer et al. (2015) manipulate personalization based on gender, but do not report a manipulation check for perceived personalization. Maslowska et al. (2016) report that using gender and occupation

(i.c., being a student or not) combined made people feel more personally addressed than a generic ad in the pretest, but do not find a significant effect on perceived personalization in their main study. As mentioned, Zarouali et al. (2018) report a significant increase in perceived personalization when adding information about both respondents’ age, gender, and location combined. It is unclear, however, whether gender in itself adds to this result. Other studies have also used gender as a personalization element combined with other personalization elements (Brinson & Eastin, 2016; Higgins et al., 2018;

Kaspar et al., 2019; Walrave et al., 2016), but do not report whether their manipulation results in differences in perceived personalization perception.

Higgins et al. (2018) use age as a separate personalization element, as well as combined with gender.

However, they do not report a manipulation check. Their findings indicate that personalization based on either the respondent’s age or gender or the combination of the two personalization elements (age and

32 Chapter 2: Let’s get personal gender) increases click-through rates. This increase in click-through rate might suggest that age, gender and the combination of both are successful in eliciting perceived personalization. Other studies use age in combination with other personalization elements as well (e.g., Kaspar et al., 2019: in combination with gender, professional interest, place of residence and current occupation; Walrave et al., 2016: in combination with first name, gender, interests and profile picture and in combination with first name, gender, interests, surname, address, email address, album picture and status update).

Song et al. (2016) expose participants to one of two scenarios, either one in which they read an e-mail containing only a fictitious name or one in which they were proposed a “personalized mortgage plan” based on their demographic characteristics (i.e., age and occupation) and other personal information that supposedly appeared in their personal Tweets. Their manipulation check shows a significant 2 difference in perceived personalization between the two scenarios. Zarouali et al. (2018) successfully use age combined with gender and location in their manipulation of personalization.

People also go through different life stages, for example going to college, graduating, moving or getting married. Reported life events on social media mark transitions between different life stages. These life stages present opportunities for marketers to target people (Kotler & Keller, 2006). Figure 2.4. shows an example from Walmart targeting students going back to college and Figure 2.5. shows a Facebook advertisement from Pampers targeted at young parents.

33 Chapter 2: Let’s get personal

Figure 2.4. Facebook advertisement using a life-event by Walmart

Note. Reprinted from Facebook Ad Library, 2019. Retrieved from https://www.facebook.com/ads/library/?active_status=all&ad_type=all&country=US&q=walmart

Figure 2.5. Facebook advertisement using a life-event by Pampers

Note. Reprinted from Facebook Ad Library, 2019. Retrieved from https://www.facebook.com/ads/library/? view_all_page_id=225923090802904 On Facebook, Google, and YouTube, advertisers can use reported life events to target consumers. For example, a company selling wedding rings is likely to target people who just got engaged. To the best

34 Chapter 2: Let’s get personal of our knowledge, life stages (or life events) have not been used in research on personalized advertising in general or on social networking sites in particular. Nevertheless, personalization based on life events bears great value for advertisers as it can provide an excellent reason to engage with a consumer (Bedgood, 2017). According to Think With Google (2017), using life-stage marketing increased Origin

Energy’s, an Australian energy provider, ad recall by 36% and searches for Origin Energy with 60%. This could suggest that life events positively contribute to the perception of personalization.

Interests Many advertisers and researchers use a person’s interests as a base for advertising personalization.

Figure 2.6., for example, shows an example from Krachttraining Voor Vrouw targeting women who are interested in power training to stay power women and to strengthen their power (which is based on 2 interactions with other Facebook pages).

Figure 2.6. Facebook advertisement using interests by Krachttraining Voor Vrouwen

Note. Reprinted from Facebook Ad Library, 2019. Retrieved from https://www.facebook.com/ads/library/?active_status=all&ad_type=all&country=ALL&q=krachttraining%20voor%20vrouwen For example, Kalyanaraman and Sundar (2006) use three levels of personalization in which they use a higher number of reported interests (e.g., sports, movies, and books) to personalize the content shown on MyYahoo.com (e.g., bookmarks to other Websites, sports news headlines, news clipper). Using more interests increases the level of perceived personalization. Aguirre et al. (2015) asked respondents to

35 Chapter 2: Let’s get personal imagine that they needed a car loan to purchase a new car and that they had sent a message to a friend on Facebook to ask for information about car loans. Then they were redirected to a Facebook homepage which contained an advertisement for a home loan (slightly personalized) or a car loan (more personalized). They report that the perceived level of personalization is higher in the more personalized condition than in the slightly personalized condition. Li (2016) reports that ads based on people’s preferred travel destination increase perceived personalization. Karmakar and Webster (2018) show a message about how increasing pollution increases allergies and breathing problem to those who indicated that allergies were a critical environmental issue for them. This personalization based on respondents’ environmental interests results in higher perceived personalization. Kim and Gambino

(2016) report that respondents who viewed a personalized advertisement for a restaurant with their preferred cuisine perceive the advertisement as being more personalized than an advertisement for a restaurant with a randomly selected cuisine. However, the personalization condition combines the preferred cuisine with the label “Foodies.com choice for YOU,” which might have also contributed to the perception of personalization. Several other studies manipulate personalization by using interests, either in itself (Sundar & Marathe, 2010) or in combination with other personalization elements (i.e., most viewed brand, Bleier & Eisenbeiss, 2015; i.e., education, Tucker, 2014; e.g., first name, gender, birthday, profile picture, Walrave et al., 2016) without reporting whether this has the intended effect on perceived personalization.

Friend referrals

Finally, marketers tap into normative perceptions by pointing out that (one or more of) a person’s friends like the advertising company. For example, Figure 2.7., shows an advertisment from Centerparcs with reference to 17 other friends who liked Centerparcs.

36 Chapter 2: Let’s get personal

Figure 2.7. Facebook advertisement using friend referrals by Centerparcs

Note. Reprinted from Facebook Ad Library, 2019. Retrieved from https://www.facebook.com/ads/library/?active_status=all&ad_type=all&country=ALL&q=center%20parcs%20belgium

Kim et al.’s (2015) survey shows that perceptions about other users’ evaluations or opinions (i.e., friends liking the advertiser) influence users’ normative perceptions. Using the specific names of a consumer’s friends might be important in signaling perceived personalization because friends are highly idiosyncratic. Windels et al. (2018) examine the visual attention participants pay to an advertisement with and without social context (i.e., the mention of friend referrals). They propose that adding social context increases personalization. While they do not report a manipulation check on perceived personalization, their finding that participants paid attention to ads without social context longer is counterintuitive and raises questions on whether social context ads actually induce perceived personalization.

The relative importance of personalization elements

As demonstrated above, different researchers have used different personalization elements (in different combinations) to manipulate personalization. While some personalization elements (e.g., location, age, gender, life events) are more general and shared by a larger group of customers, others (e.g., interests or friend referrals) are more idiosyncratic to a specific customer or a smaller segment. Simonson (2005)

37 Chapter 2: Let’s get personal suggests that more idiosyncratic elements will, in general, be overweighed in consumers’ evaluations of product information as they distinguish consumers more from other consumers. However, there is no empirical evidence that we know of that could support this claim. Windels et al. (2018), for example, report that friend referrals, which are idiosyncratic, do not generate more visual attention. This finding might suggest that friend referrals do not contribute strongly to perceived personalization. All in all, the evidence from previous research is insufficiently clear how strongly each variable will contribute to the perception of personalization. Therefore, we formulate the following research question:

RQ1: What is the relative importance of using location, age, gender, life events, interests, and friend referrals as personalization elements in an advertisement on a social networking site in evoking perceived personalization of the advertisement?

The effect of perceptions about product characteristics

Different personalization elements might not be equally important for the evaluation of perceived personalization for different individuals, depending on their perceptions of products or services as being more or less involving, possessing search or experience qualities, and their hedonic or utilitarian buying motivations.

Product category involvement

First, the relative importance of personalization elements could differ between lowly and highly involved individuals. According to the Elaboration Likelihood Model (ELM) (Petty & Cacioppo, 1986), high product category involvement causes more elaborate (central) information processing, whereas low product category involvement leads to peripheral processing.

As argued above, more idiosyncratic elements (i.e., personal interests or friend referrals) are more likely to evoke perceived personalization than more common characteristics. This could especially hold for highly involved individuals that process the advertisement more elaborately. They are more likely to scrutinize the different personalization elements and may find that only an ad appealing to truly idiosyncratic elements are genuinely personalized. In contrast, more lowly involved consumers might consider any personalization elements (e.g., gender, age, location, life events) as a sufficient cue to signal personalization because they process the advertisement more superficially. Because there is no prior empirical evidence on this matter, we propose the following research question:

38 Chapter 2: Let’s get personal

RQ2: How does the relative importance of using location, age, gender, life events, interests, and friend referrals as personalization elements in an advertisement on a social networking site differ in evoking perceived personalization for highly involved compared to lowly involved consumers?

Search and experience qualities People can perceive goods and services as possessing primarily search or experience qualities (Nelson,

1970; Zeithaml, 1981). Appliances or furniture, for example, can be evaluated pre-purchase and are, therefore, perceived as having primarily search qualities (Xia & Bechwati, 2008; Zeithaml, 1981). A restaurant or theatre show, on the other hand, can only be evaluated post-purchase (by actually experiencing the product) and are, therefore, often perceived as having primarily experience qualities

(Hsieh, Chiu, & Chiang, 2005; Zeithaml, 1981). Experiences are more personal and more subject to 2 personal taste. The preferences for experience products may, therefore, be more heterogeneous than for search products. Feick and Higie (1992) show that in the case of high preference heterogeneity, consumers will focus attention on similarity information (in their case, with an endorser) and similarity will be an important determinant of influence. More idiosyncratic personalization elements (such as the fact that your friends like this product) might signal a greater deal of similarity with the intended target group, and therefore a greater degree of perceived personalization, than more general characteristics

(such as your location). Search goods, on the other hand, may be more standardized and more a matter of “opinion”: claims can at least in principle be evaluated at some level with regard to their correctness. People may be happy that they share their opinions with others, as this attests to their correctness

(Quaschning, 2013). They may, therefore, consider any personalization cue as sufficient to perceive an advertisement to be personalized.

To our knowledge, there is no empirical evidence on how personalization elements differently elicit perceived personalization for search and experience products and services, therefore, we propose the following research question:

RQ3: How does the relative importance of using location, age, gender, life events, interests, and friend referrals as personalization elements in an advertisement on a social networking site differ in evoking perceived personalization for experience compared to search products and services?

39 Chapter 2: Let’s get personal

Hedonic and utilitarian buying motivations

Finally, goods and services can be bought to fulfill either hedonic or utilitarian buying motivations (Batra

& Ahtola, 1991; Dhar & Wertenbroch, 2000). Services or goods that are bought for predominantly hedonic motivations (e.g., a restaurant) provide affective and sensory experiences (Hirschman &

Holbrook, 1982). As a result, hedonic services are highly person-specific (Voss et al., 2003). A utilitarian buying motivation (e.g., a bank), on the other hand, accomplishes a functional or practical task. These tasks are often less person-specific: the evaluation of the quality can be made more objectively (Voss et al., 2003). Therefore, it might not be important to use highly idiosyncratic personalization elements to evoke perceived personalization. Due to the person-specific nature of hedonic buying motivations, it can be expected that, contrary to a utilitarian motivation context, idiosyncratic personalization elements are more important to elicit perceived personalization than general characteristics. Due to a lack of empirical evidence on the matter, we propose the following research question:

RQ4: How does the relative importance of using location, age, gender, life events, interests, and friend referrals as personalization elements in an advertisement on a social networking site differ in evoking perceived personalization for consumers with hedonic compared to utilitarian buying motivations?

The effect of audience demographics

In order to be able to contribute to management decisions, we also included gender, age, and education as potential moderating variables. However, it is unclear what the impact of these three socio- demographic variables might be on the evaluation of perceived personalization. With regard to gender, the selectivity model (Meyers-Levy & Loken, 2015; Meyers-Levy & Sternthal, 1991) proposes that women tend to process information taking all the available cues into account, assigning equal importance to information relevant to themselves and to others. This could mean that even more subtle personalization cues would signal personalization to women. On the other hand, due to the fact that women are more systematic processors (Bušljeta Banks, Dens, & De Pelsmacker, 2016), it is also possible that they focus more on the idiosyncratic cues to asses perceived personalization due to the fact that only those cues truly differentiate them from other consumers. Men, in contrast, use an item- specific processing style and tend to rely on a single salient cue or a subset of cues (Bušljeta Banks et

40 Chapter 2: Let’s get personal al., 2016). As such, we would expect men to focus more on the more salient, idiosyncratic, personalization elements to infer personalization.

With respect to age, previous research suggests that especially young consumers are susceptible to social norms as a result of the consumer socialization process (Moore & Bowman, 2006). Therefore, it may be expected that for younger consumers, the referral to personal connections liking the product is a more important personalization cue than for older consumers.

Education has often been linked to the need for cognition (Cacioppo, Petty, Feinstein, & Jarvis, 1996): people with higher levels of education are often found to be in higher need for cognition (NFC). Need for cognition, in turn, results in more elaborate processing: those high in need for cognition are more likely to elaborate and generate inferences in response to advertisements than those low in need for 2 cognition (Haugtvedt, Petty, & Cacioppo, 1992; Sicilia, Ruiz, & Munuera, 2005). As a result, they might focus more on the more idiosyncratic personalization elements, compared to lower educated consumers to assess the perceived personalization because only those might be successful in differentiating them from other consumers. Due to the fact that, in general, it is unclear what the effect of these three socio- demographic characteristics might be, we pose the following research question

RQ5: What is the moderating effect of demographic characteristics of users on the relative importance of location, age, gender, life events, interests, and friend referrals in an advertisement on a social networking site in evoking perceived personalization?

Method

To determine the relative importance of various personalization elements for perceived advertising personalization, we used conjoint analysis. Conjoint analysis involves exposing respondents to a set of

(in this case, eight) advertisements (‘cards’) that systematically vary in (combinations of) personalization elements (called ‘attributes’). Based on the perceived personalization score for each advertisement, the implicit relative importance of each attribute (based on the ‘part-worth utilities’ of the levels) can be derived.

41 Chapter 2: Let’s get personal

Pretests

In conjoint analysis, the selection of attributes and levels is an important factor for the calculation of the relative importance of the attributes. To determine the attributes (the actual personalization elements), we presented 37 respondents (푋̅age= 34.65, SDage = 11.75; 48.6% male) with a list of 11 possible personalization elements (age, education, location, work, life events, gender, relationship status, interests, own page-likes, friends and friend referrals), and asked them to indicate on two seven- point Likert items from Weathers, Sharma, and Wood (2007) to what extent they found these personalization elements both useful and annoying as a basis for personalization on social media. We selected the three personalization elements that consumers perceived as the most useful for personalization (p> .054): interests (푋̅useful= 4.51, SDuseful= 1.84; 푋̅annoying= 3.19, SDannoying = 1.68), age

(푋̅useful= 4.03, SDuseful= 1.80; 푋̅annoying= 2.95, SDannoying = 1.76), and gender (푋̅useful= 4.19, SDuseful= 1.76;

푋̅annoying= 3.11, SDannoying = 1.91) and the three that they perceive as most annoying (p> .008): friend referrals (푋̅annoying= 4.95, SDannoying = 1.90; 푋̅useful= 2.65, SDuseful= 1.86), life events (푋̅annoying= 4.70,

SDannoying = 1.73; 푋̅useful= 3.24, SDuseful= 1.67) and location (푋̅annoying= 4.59, SDannoying = 1.77; 푋̅useful=

3.35, SDuseful= 1.60). The “useful” cues were considered significantly more useful than they were annoying, and vice versa.

We conducted six separate conjoint analyses for six products: a bank, a smartphone, tableware, furniture, a restaurant, and a fashion retailer. The choice for these products is based on a pretest, in which we aimed to select three search and three experience products/services with a substantial variation in involvement. We tested 29 potential products and services based on prior research. We then asked 61 respondents (Mage = 31.7541; SDage = 9.384, 55.7% male) to rate the search and experience qualities of these products and services on two seven-point Likert scales from Weathers et al. (2007).

To avoid fatigue, each respondent evaluated a random selection of 15 of the 29 products. We selected the three products and services with the highest score on experience qualities (Mrestaurant = 6.129,

SDrestaurant = 1.088; Mfurniture = 5.935, SDfurniture = 1.237; Mfashion retailer = 5.935, SDfashion retailer= 1.263) and the three with the highest score on search qualities (Msmartphone = 5.109, SDsmartphone = 1.190; Mbank: M =

5.063, SDbank = 1.287; Mtableware = 5.031, SDtableware = 1.385). The selected products all appeared to be at least somewhat involving. People are not very likely to engage with advertisements they are not at least minimally involved with (Msmartphone = 6.083, SDsmartphone = 1.023; Mbank = 5.906, SDbank = 1.024;

42 Chapter 2: Let’s get personal

Mfashion retailer = 5.602, SDfashion retailer = .960; Mrestaurant = 5.591, SDrestaurant = .922; Mfurniture = 5.237, SDfurniture

= 1.206.; Mtableware = 4.417, SDtableware = 1.515).

Main study design To develop the conjoint stimuli for the main study, the six attributes selected from the pretest were each manipulated on two levels: either the advertising message referred to the personalization element or the element was omitted (in other words, a generic version was used). We used SPSS orthoplan to create a balanced orthogonal design of eight advertisements (cards) (Table 2.2.). We employed traditional full-profile conjoint analysis, where respondents needed to score all eight cards (Hair, Black,

Babin, Anderson, & Tatham, 2010). Each card consists of a different combination of levels and attributes, but each attribute level appears an equal number of times across the eight stimuli (De Meulenaer, Dens, 2 & De Pelsmacker, 2015). For example, the attribute “age” as a personalization element appears in four out of eight cards (the other four cards are not personalized based on age). Of these four cards, two also contain “gender” as a personalization element, while the other two do not personalize on gender.

Table 2.2. Orthoplan

Attributes Stimulus Age Gender Location Life events Interests Friend referral 1 Not Not Mentioned Mentioned Not Mentioned mentioned mentioned mentioned 2 Mentioned Mentioned Mentioned Not Not Mentioned mentioned mentioned 3 Mentioned Not Not Mentioned Mentioned Mentioned mentioned mentioned 4 Not Not Mentioned Not Mentioned Not mentioned mentioned mentioned mentioned 5 Mentioned Not Not Not Not Not mentioned mentioned mentioned mentioned mentioned 6 Mentioned Mentioned Mentioned Mentioned Mentioned Not mentioned 7 Not Mentioned Not Not Mentioned Mentioned mentioned mentioned mentioned 8 Not Mentioned Not Mentioned Not Not mentioned mentioned mentioned mentioned

To manipulate “age”, “gender”, “location”, “interests” and “life events”, we inserted respondents’ data based on the information they provided at the start of the questionnaire (see Table 2.3.). For friend referrals, we used a generic sentence (“3 of your friends like …”) rather than asking them to write down actual friend names, to avoid potential confounds caused by the relationship with specific friends (e.g., in terms of homophily, tie strength or source credibility).

43 Chapter 2: Let’s get personal

Table 2.3. Manipulations per attribute

Attribute Manipulation mentioned Location Headline: “The #1 choice (product) for people near you in (state)”, Age Headline: “The #1 choice (product) for people under 35” Gender Headline: “The #1 choice (product) for (men/women)” Interests Headline: “The #1 choice (product) for people that are interested in (socializing)” Friend referrals Additional statement above the advertisement: “3 of your friends like (brand’s name)” Life event Additional statement above the advertisement: “Other users who recently (life event) like ’(brand’s name).”

Each advertisement was introduced by a short scenario explicating the personalization elements that were used. For example, “Imagine that you would encounter the advertisement (suggested post) below in your Facebook newsfeed, based on your location, life events and the page-likes from your friends.”

For location, the advertisement headline read “The #1 choice (product) for people near you in (home state)”, When the advertisement was personalized based on age, the text above the picture read “The

#1 choice (product) for people under 35”, because we only selected respondents between the ages of

18 and 35. For gender, the advertisements read “The #1 choice (product) for (men/women)”. The same was done for interests, for example, “The #1 choice (product) for people that are interested in

(socializing)”. For life events, we included a statement above the advertisement “Other users who recently (life event) like (brand name).” For the friend referrals, we inserted the text “3 of your friends like (brand’s name)” above the advertisement. Depending on the selected attributes, combinations of the above manipulations were used: for example “The #1 choice (product) for (men/women) under 35 near you in (state)”. The picture in the advertisement and the rest of the advertising copy (call to action) were kept constant over the different advertisements for each product or service (an example is shown in Appendix A.1.). For each of the 8 advertisements, respondents rated the perceived personalization of the advertisement (8 items, αminimum = .934, based on Kalyanaraman and Sundar (2006); Maslowska et al. (2016); Srinivasan, Anderson, and Ponnavolu (2002)) on a seven-point Likert scale. Construct scores per advertisement were computed by calculating the average of the items per advertisement for use in further analyses. See Table 2.4. for an overview of the construct items.

44 Chapter 2: Let’s get personal

Table 2.4. Measures

Construct Items Cronbach’s alpha Median Interests Socializing Traveling Sports Golf Art and Culture Camping Music Hiking Skiing Reading Gaming Nature Shopping Going out Life events Got a new job Started at a new school Studied abroad Got a new relationship Got engaged Got married 2 Celebrated an anniversary Got a new child Moved Bought a home Did home improvement Traveled None of the above Perceived 1) This ad is tailored to my situation. αadvertisement 1 = .938 αadvertisement 2 = personalization 2) I believe this ad is customized to my .942 αadvertisement 3 = .945 αadvertisement 4 needs. = .946 αadvertisement 5 = .934 3) This ad was targeted at me as a αadvertisement 6 = .944 αadvertisement 7 = unique individual. .941 αadvertisement 8 = .936 4) I believe that this ad is customized to my characteristics. 5) This ad was personalized according to my profile. 6) There was personal information in the ad. 7) The ad was targeted at me. 8) I could recognize myself in the group the ad was targeted at. Product category [Product/service] to me is .944 involvement 1) unimportant – important 2) meaningless – meaningful 3) does not matter to me – does matter to me Search product 1) I can adequately evaluate .807 4.00 qualities [product/service] using only the information provided by the retailer about its characteristics. 2) I can evaluate the quality of [product/service] simply by reading information about [product/service]. Experience It’s important for me to personally [try 6.00 product qualities tableware/ a bank/ a sofa / on clothing / a smartphone / a restaurant] to evaluate it. Hedonic buying If I were to [buy tableware / a sofa / .895 5.33 motivation clothing / a smartphone / visit a restaurant / choose a bank], I would primarily do this because it is …

45 Chapter 2: Let’s get personal

1) not fun – fun 2) dull – exciting 3) unenjoyable - enjoyable Utilitarian buying If I were to [buy tableware / a sofa / .876 6.33 motivation clothing / a smartphone / visit a restaurant / choose a bank], I would primarily do this because it is … 1) ineffective – effective 2) not functional – functional 3) impractical - practical Gender Are you 1) Male 2) Female Age Calculated based on birth year: 27.00 What year were you born in? Education What is the highest level of education 5.00 you have completed? If you are a student, please indicate your current level of education. 1) Primary school 2) Middle (junior high) school 3) High school 4) Bachelor’s degree (Undergraduate) 5) Master’s degree (Graduate) 6) Ph.D. (doctorate) or postgraduate Days spent on In the past week, on average, FB approximately, how many days have you used Facebook? 1) 1 day 2) 2 days 3) 3 days 4) 4 days 5) 5 days 6) 6 days 7) 7 days Minutes per day In the past week, on average, how spent on FB many minutes per day have you spent on Facebook? 1) less than 10 minutes 2) between 11 and 30 minutes 3) between 31 and 60 minutes 4) more than 1 hour, but less than 2 hours 5) more than 2 hours, but less than 3 hours 6) more than 3 hours Data collection procedure and sample

Using the crowdsourcing website Prolific, we selected panel members that use Facebook at least once a month. Panel members are usually motivated to participate and have experience with online surveys.

Kees, Berry, Burton, and Sheehan (2017) have offered evidence that the results from a crowdsourcing website (in their study, MTurk) outperformed panel data from a professional marketing research company and performed comparably to student samples in terms of data quality. A quality check consisted of excluding partially completed questionnaires and rejecting respondents who rated all advertisements the same or inconsistently responded to reverse-scaled items (De Meulenaer et al.,

46 Chapter 2: Let’s get personal

2015). The final sample consisted of 595 U.S. respondents (Mage=27.87, SDage=3.93; 46.7% male).

Most of the respondents had at least a graduate degree (70%). 53.8% of the respondents were active on Facebook on a daily basis, most spent less than one hour per day on Facebook (72.1%).

The survey started with a welcome screen informing participants about the general aim of the study and guaranteeing anonymity. Respondents were first asked to indicate which social media they used at least once a month to verify the pre-screening. Next, socio-demographic information was collected (gender, year of birth, education, home state). Then, respondents were asked to choose the topics they were most interested in out of a list of 14 possibilities and to select a life-event that had happened to them in the last six months out of a list of 12 possibilities (see Table 2.4. for the complete lists). On the next page, we asked them about their Facebook use (amount of days and minutes per day). Then, 2 respondents were sent to a short instruction page in which they were informed that they would be presented with 8 different advertisements that they might encounter on Facebook. They were asked to look at the advertisements attentively. They were informed that the pages might look highly similar but were subtly different. Then, they were asked to rate each advertisement individually, disregarding the evaluations of any previous advertisement. The order of the advertisements was randomized to avoid confounds due to learning or wear-out effects.

Analyses and Results

SPSS conjoint was used to carry out the analyses. For each respondent, the relative importance of each attribute (personalization element) is calculated based on the part-worth utilities for the two levels making up the attribute. Attributes with greater part-worth ranges are considered to be more important than those with smaller ranges. The importance percentages are computed by taking the utility range for each respondent for each attribute and dividing it by the sum of the utility ranges for all attributes. Afterward, the average percentage over the respondents is calculated.

First, we pooled the different products and services across the six products to answer RQ1 (Table 2.5.).

All part-worths are positive which indicates that each personalization element increases the perception of personalization. The most important elements to evoke perceived personalization are interests

(20.953%), location (18.595%) and age (17.592%). The least important elements are gender

(15.426%), friend referrals (13.759%) and life events (13.675%).

47 Chapter 2: Let’s get personal

Table 2.5. Conjoint analysis for perceived personalization: pooled across products

Attribute Level Overall Part-worth Importance Location Not mentioned -.168 18.595 Mentioned .168 Age Not mentioned -.168 17.592 Mentioned .168 Gender Not mentioned -.077 15.426 Mentioned .077 Life-event Not mentioned -.016 13.675 Mentioned .016 Interests Not mentioned -.215 20.953

Mentioned .215 Friend referrals Not mentioned -.076 13.759 Mentioned .076 Constant 4.244

Next, we median split (a separate split for each variable) the sample based on consumers’ perceptions of the different product characteristics (product category involvement, search vs experience qualities and utilitarian vs hedonic buying motivations). We then reran the conjoint analyses for the separate samples. Table 2.6. shows the part-worths for each attribute level for each median-split product characteristic. To answer RQ2, we compared the results for the two samples differing in product category involvement. For both less and more involved consumers, each personalization element increases the perception of personalization. The results are highly comparable between the two groups and follow the pattern described above. For both groups, interests (low involvement: 20.907% versus high involvement: 21.340%) contribute the most to perceived personalization, followed by location

(18.287% versus 18.703%), age (17.479% versus 16.860%), and gender (15.133% versus 15.507%).

For both groups, the elements that contribute the least are friend referrals (low involvement: 13.525% versus high involvement: 14.241%) and life events (14.669% versus 13.349%), although the order is reversed: Life events are less important than friend referrals for more involved consumers, while the reverse is true for less involved individuals. The differences are very small, though.

To answer RQ3, we compared whether there were differences between products and services that were evaluated as having high search (i.e., above the median on the “search” scale) versus high experience

(i.e., above the median on the “experience” measure) qualities. Consumers scoring above the median on both scales (n = 69) were excluded from the analysis to avoid counting them twice. For both types of products, each personalization element increases the perception of personalization. The results are

48 Chapter 2: Let’s get personal again highly comparable between the two types of products, and, in general, follow the pattern described above. For both groups, interests (search: 20.799% versus experience: 21.088%) contribute the most to perceived personalization. Location (search: 18.624% versus experience: 17.514%) and age (search: 18.073% versus experience: 18.556%) take up the second and third place, with location being the second most important element for search products and services, and age for experience products and services. For both groups, the elements that contribute the least to perceived personalization are gender (search: 14.988% and experience: 13.225%), friend referrals (search:

14.413% versus experience: 15.846%) and life events (search: 13.104% versus experience: 13.770%). While the ranking of the elements differs slightly between

50 Table 2.6. Conjoint analysis for perceived personalization: products pooled across product characteristics Let 2: Chapter

Attribute Level Low involvement High involvement Search Experience Hedonic Utilitarian Part- Importance Part- Importance Part- Importance Part- Importance Part- Importance Part- Importance worth worth worth worth worth worth Location Not -.167 -.167 -.180 -.176 -.182 -.155

’ mentioned 18.287 18.703 18.624 17.514 18.936 17.620 personal s get Mentioned .167 .167 .180 .176 .182 .155 Age Not -.187 -.134 -.182 -.178 -.154 -.183 mentioned 17.479 16.860 18.073 18.556 17.246 17.884 Mentioned .187 .134 .182 .178 .154 .183 Gender Not -.089 -.060 -.055 -.034 -.043 -.101 mentioned 15.133 15.507 14.988 13.225 14.093 15.371 Mentioned .089 .060 .0557 .034 .043 .101 Life-event Not -.011 -.019 -.035 -.045 -.009 -.023 mentioned 14.669 13.349 13.104 13.770 15.103 13.233 Mentioned .011 .019 .035 .045 .009 .023 Interests Not -.185 -.245 -.193 -.240 -.222 -.246 mentioned 20.907 21.340 20.799 21.088 21.710 22.600 Mentioned .185 .245 .193223 .240 .222 .246 Friend Not -.083 -.073 -.058 -.122 -.068 -.085 referrals mentioned 13.525 14.241 14.413 15.846 12.911 13.291 Mentioned .083 .073 .058 .122 .068 .085 Constant 3.947 4.512 4.413 4.121 4.118 4.342

Notes: Rlow involvement = .990, p < .001; Rhigh involvement = .999, p < .001; Rsearch = .995, p < .001; Rexperience = .987, p < .001; Rhedonic = 1.000, p < .001; Rutilitarian = .997, p < .001

Chapter 2: Let’s get personal

the two product types (For search products, life events are the least important, while it is gender for experience products, the differences are very small.

Third, we compared between products and services that were rated as either being highly hedonic (above the median) or being highly utilitarian (above the median), to answer RQ4. As in the previous analysis, we excluded consumers who rated the product or service high on both hedonic and utilitarian buying motivations (n = 127). Again, all personalization elements contribute positively to the perception of personalization. For both utilitarian and hedonic products or services, we find that interests (hedonic: 21.710% versus utilitarian: 22.600%), location (hedonic: 18.936% versus utilitarian: 17.620%) and age

(hedonic: 17.246% versus utilitarian: 17.884%) contribute most to perceived personalization, whereas gender (hedonic: 14.093% versus utilitarian: 15.371%), life events (hedonic: 15.103% versus 2 13.233%) and friend referrals (hedonic: 12.911% versus utilitarian: 13.291%) contribute least to perceived personalization. While there are small differences in the relative importance and the ranking between the two groups (for example, gender is relatively more important for utilitarian products, where personal get Let’s 2: Chapter life events are relatively more important for hedonic products), the differences are smaller than 2%. We, therefore, conclude that the results are highly stable across products. Finally, we compared the results between respondents with different demographic characteristics to answer RQ5 (Table 2.7.).

First, we compared the importance values between men and women. The pattern for both men and

women follows the general pattern described in the previous analyses: interests (men 20.969% versus women: 20.934%), location (men: 19.176% versus women: 17.176%) and age (men: 17.699% versus women: 17.471%) contribute most to perceived personalization, whereas gender (men: 14.766% versus women: 16.173%), friend referrals (men: 13.784% versus women: 13.730%) and life events (men: 13.606% versus women: 13.753%) contribute the least. For men, location contributes more to perceived personalization compared to women and women attribute a higher importance to gender. However, these differences are again very small (less than 2%).

Table 2.7. Conjoint analysis for perceived personalization: pooled across demographic characteristics personal get Let’s 2: Chapter

Attribute Level Men Women Age: under 27 Age: over 27 Education: low Education: high Part- Importance Part- Importance Part- Importance Part- Importance Part- Importance Part- Importance worth worth worth worth worth worth Location Not -.170 19.176 -.166 17.938 -.215 19.776 -.116 17.467 -.170 19.177 -.161 17.503 mentioned Mentioned .170 .166 .215 .116 .170 .161 Age Not .164 17.699 -.172 17.471 -.121 17.370 -.198 17.910 -.142 17.477 -.149 18.525 mentioned Mentioned .164 .172 .121 .198 .142 .149 Gender Not -.057 14.766 -.099 16.173 -.065 15.299 -.076 15.675 -.051 15.738 -.050 14.907 mentioned Mentioned .057 .099 .065 .076 .051 .050 Life-event Not -.015 13.606 -.018 13.753 -.042 13.518 .004 13.324 -.012 13.618 -.014 13.131 mentioned Mentioned .015 .018 .042 -.004 .012 .014 Interests Not -.210 20.969 -.221 20.934 -.243 21.384 -.175 20.473 -.220 20.355 -.144 21.108 mentioned Mentioned .210 .221 .243 .175 .220 .144 Friend Not -.081 13.784 -.071 13.730 -.061 12.653 -.092 15.151 -.069 13.636 -.092 14.826 referrals mentioned Mentioned .081 .071 .061 .092 .069 .092 Constant 4.199 4.296 4.195 4.292 4.091 4.363

Notes: Rmen = .998, p < .001; Rwomen = .998, p < .001; Runder 27 = .995, p < .001; Rover 27 = .999, p < .001; Rlow education= .996, p < .001; Rhigh education = .997, p <

.001

Chapter 2: Let’s get personal

Next, we compared respondents 18-27 to those 28-35 years old (based on a median split of the sample).

Interests (younger respondents: 21.384% versus older respondents: 20.473%) and age (younger respondents: 17.370% versus older respondents: 17.910%) contribute most to perceived personalization for both groups, whereas life events (younger respondents: 13.518% versus older respondents: 13.324%) and gender (younger respondents: 15.299% and older respondents: 15.675%) contribute the least to perceived personalization for both groups. Location is more important for younger respondents (19.776%) than for older respondents (17.467%), while friend referrals weigh more for older (15.151%) than for younger respondents (12.653%).

Finally, we compared lower to higher educated respondents. Again, the same pattern as described above emerges. Interests (lower educated respondents: 20.355% versus higher educated respondents: 2 21.108%), location (lower educated respondents: 19.177% versus higher educated respondents: 17.503%) and age (lower educated respondents: 17.477% versus higher educated respondents:

18.525%) contribute most to perceived personalization. Gender (lower educated respondents: 15.738% personal get Let’s 2: Chapter versus higher educated respondents: 14.907%), friend referrals (lower educated respondents: 13.636% versus higher educated respondents: 14.826%) and life events (lower educated respondents: 13.618% versus higher educated respondents: 13.131%) are least important. While there are some minor differences (location is more important for lower than for higher educated respondents, whereas age

and friend referrals are more important for higher educated than for lower educated respondents), they don’t exceed 2%.

Discussion

The present study is, to our knowledge, the first to investigate the relative importance of antecedents of perceived personalization for products and services with different product characteristics and for different target groups. Since previous literature indicates that perceived personalization and actual personalization do not necessarily match (De Keyzer et al., 2015; Kramer et al., 2007; Li, 2016), it is crucial for both academics and managers to understand how the perception of personalization could be achieved, and which elements contribute most to this perception. We examined six personalization elements that can be used on Facebook. Nevertheless, these personalization options are not limited to

Facebook. On Instagram (2019), advertisers can also use location, demographics, interests and behavior

53 Chapter 2: Let’s get personal as personalization elements. Twitter (2019) reports that you can use more or less the same options

(e.g., location, gender, interests, behavior, and followers) and also Pinterest (2019) allows advertisers to personalize messages based on age, location, gender, interests, and keywords users searched for.

On LinkedIn, it is mandatory to specify location. Other personalization elements on LinkedIn are company information, demographics, education, job experience and interests.

The findings of the current study indicate that all elements positively contribute to the perception of personalization with interests contributing the most. Prior studies by Kalyanaraman and Sundar (2006) and Li (2016), for example, manipulating personalization through interests also report that their manipulations were successful in eliciting perceived. Location and age also contribute effectively to perceived personalization, consistent with the results of Zarouali et al. (2018). Friend referrals and life events contribute least to perceived personalization. Although both elements contribute positively to perceived personalization, our findings indicate that they are relatively less effective in triggering perceived personalization than the other elements. These two elements were also among the least studied in prior research.

Our findings indicate that the relative importance of personalization elements is relatively stable across product characteristics and demographic characteristics of the respondent. This indicates that, when trying to manipulate perceived personalization, be it for research or for marketing purposes, interests, location, and age are highly likely to have the strongest impact on perceived personalization, across product categories and demographic market segments.

Managerial implications

Previous research already indicated that, when an advertiser wants to create a personalized advertisement, it is important to choose those personalization elements that indeed elicit perceived personalization. Our findings provide guidelines on which elements should be used. Advertisers wanting to evoke perceptions of personalized advertisements should primarily use a consumer’s interests, as this is the strongest determinant of perceived personalization. Friend referrals and life events seem less effective. As each of the personalization elements contributes positively to perceived personalization, managers could consider combining all possible personalization elements to induce the highest possible level of perceived personalization. Previous research, indeed, has often combined different

54 Chapter 2: Let’s get personal personalization elements to manipulate higher levels of personalization (e.g., De Keyzer, Kruikemeier, et al., 2019; Walrave et al., 2016; Zarouali et al., 2018). Our research indicates that this might be a good strategy, as there should be a cumulative effect when adding multiple elements. However, managers should weigh the costs and benefits. More narrow targeting could be more costly. Paying more to personalize elements that contribute less to perceived personalization, could not be cost- effective.

Advertisers should also consider their brand effects. While perceived personalization is often considered to entail positive effects, the use of multiple personalization elements could also be considered as creepy and, therefore, results in less positive brand attitudes (De Keyzer, Kruikemeier, et al., 2019). Some specific personalization elements may also annoy consumers and should, therefore, be used more 2 cautiously. In our pretest, respondents stated to find location, life events and friend referrals as more annoying than useful. Especially since the latter two only faintly contribute to perceived personalization, it would be better to avoid using them. Nevertheless, although friend referrals are considered as a personal get Let’s 2: Chapter targeting option in social networking sites (e.g., Facebook, 2008), they might rely on a different kind of tactic, namely the use of social approval. They might not be very important to elicit perceptions of personalization, however they might affect consumers’ responses via perceptions of social approval.

Limitations and Suggestions for Future Research

Limitations of the present study provide opportunities for future research. First of all, we have used only six personalization elements. While these correspond to elements that are frequently used in advertising on social networking sites, prior research has also included other elements (such as the respondent’s name or work status) (e.g., Li & Liu, 2017; Maslowska et al., 2016). These could also be tested in order to determine their relative importance on perceived personalization. Future research could also examine other levels for each of the personalization attributes. For example, the current study manipulated friend referrals by stating that three of the respondents’ friends liked the page. Indicating that more than three friends could have a different effect on perceived personalization. Moreover, it might be more important to show consumers that are highly homophilious with the respondent as these friends can be expected to have similar needs and preferences according to social comparison theory (Festinger, 1954).

55 Chapter 2: Let’s get personal

Nevertheless, the current research has opted not to do this in order to prevent confounds with regard to the relationship between the sender and his friends.

Second, we manipulated the personalization elements very overtly (by using it in the slogan and in the scenario instructions) to increase the internal validity of the study. The examples above provide evidence that explicit references to personalization cues reflect reality in at least some occasions. However, personalization techniques can also be used more covertly. For example, life-events are not always used to personalize in the same overt way as we manipulated it. As a result, the results of our study might not be generalizable to more covert personalization techniques. When using covert personalization techniques, it might be harder for respondents to recognize the advertisement was personalized.

Respondents were also exposed to static advertisements and which were shown not shown in the context of a Facebook homepage, in which other information is competing for respondents’ attention.

This lack of interactivity and context could have influenced the results. As with any research, there is a tradeoff between internal and external validity. A field experiment would provide insights on the generalizability of our findings.

Third, we only examined the main effects of the six personalization elements. It remains unclear which combinations would impact the perception of personalization the most and how combinations should be built to elicit different levels of perceived personalization. Unfortunately, our research does not allow to study potential interaction effects, that could indicate synergies between elements. It is also possible that adding a less important personalization element, such as life events, to a highly important element, such as interests, does not make a difference. On the other hand, it will be important to keep in mind which personalization elements are perceived as useful for respondents. For example, in our first pretest respondents evaluated the use of location in advertising as more annoying than useful. Even though it contributes strongly to perceived personalization, it might result in negative consumer responses due to the fact that consumers find it annoying. Interests and age, on the other hand, were rated as more useful than annoying and as a result, might result in positive consumer responses.

Fourth, friend referrals were only manipulated by stating the number of friends who liked the page to avoid potential confounds with interpersonal relationships. However, as Agarwal, Lee, and Whinston

(2017) posit, rather than the overall number of friend referrals driving effects on consumer responses,

56 Chapter 2: Let’s get personal such as click-through, the relational characteristics with these friends might be more important. For example, they show that more homophilous friends, who share for example age, gender, interests, positively affect the effect of friend referrals on click intention. Future research could examine the impact of interpersonal relationships on the effects of friend referrals both on perceived personalization as on consumers’ attitudes and behavior(al intentions).

Next, we have only examined the impact of advertising elements on perceived personalization and not on attitudinal or behavioral measures. Our pretest indicated that, for example, location was perceived to be more annoying than useful and it is one of the elements that elicits perceived personalization the most. As a result, it might elicit negative personalization effects and, for example, decrease brand attitudes or purchase intentions. Future research could also examine other dependent variables to 2 disentangle the positive and negative personalization effects.

Finally, we selected products and services that were at least moderately involving, which might explain

Chapter 2: Let’s get personal get Let’s 2: Chapter the lack of differences between highly and lowly involved consumers. Although there was still adequate variance in the involvement scores for the different products and services, future research could select different products and/or services with a larger variation in involvement to examine this further.

Is this for me? How Consumers Respond to Personalized Advertising on Social Network Sites1,2

1 Manuscript published as De Keyzer, F., Dens, N., & De Pelsmacker, P. (2015). Is this for me? How Consumers Respond to Personalized Advertising on Social Network Sites. Journal of Interactive Advertising, 15(2), 124-134. 2 An earlier version of this chapter was presented at the 2014 ICORIA conference in Amsterdam, The Netherlands. 3 Chapter 3: Is this for me?

Abstract

We study the impact of perceived personalization on consumer responses to advertising on Facebook, a popular social network site. Based on two experiments, we test a moderated mediation model with perceived relevance as the mediator and respondents’ attitude toward Facebook as the moderator of the relationship between perceived personalization on the one hand and brand attitude and click intention on the other hand. The results show that perceived personalization improves responses toward

Facebook ads, through perceived relevance. The moderating impact of the attitude toward Facebook is only significant in the second study. There, the positive effect of perceived personalization of Facebook advertisements on click intention is stronger for participants with a more positive attitude toward

Facebook.

60 Chapter 3: Is this for me?

Introduction

With consumers’ increasing use of social network sites, such as Facebook, Twitter, and YouTube, advertising on these platforms has also increased (Kelly et al., 2010; Kwon & Sung, 2011). For example, two out of three Americans use social media sites, which represents about 147.8 million people (Park,

Rodgers, & Stemmle, 2011). Especially for young people, social media have become perhaps the most popular communication channels (Chu, 2011). Facebook boasts an average of 864 million daily active users worldwide (Facebook, 2014). Research suggests that social network sites can be an interesting platform for firms to reach their target group: firms are able to target specific consumer groups at a lower cost and with higher speed (Saxena & Khanna, 2013; Trusov et al., 2009; Wen et al., 2009). Globally, advertising in social media reached about $9 billion in 2013 (Magna Global, 2013). In the US alone, social advertising revenues are expected to grow from $4.7 billion in 2012 to $11 billion in 2017, representing a compound annual growth rate of 18.6% (Stadd, 2014). With this rapid growth over a 3 short period of time, academic research on social networking sites has struggled to keep pace (Kelly et al., 2010). Relatively little is known with respect to how consumers respond to advertisements on these sites, and which factors might influence consumers’ responses (Chu, 2011; 2011; Tucker, 2014; Wen et al., 2009). In the present study, we try to partly close this knowledge gap by experimentally testing the responses to advertising in a social network site, taking ad personalization, perceived relevance and the attitude towards the site into account (Figure 3.1.). The study is set on Facebook because that is considered by many as the most popular social network site today. For example, in the US, Facebook accounts for 45.03 percent of all social media site visits (Statista Inc., 2019).

One of the advantages of advertising in online social network sites is that advertising messages can be sent to specific targets on the basis of their disclosed interests and demographics (Kelly et al., 2010;

Sundar & Marathe, 2010). Personalized advertising can be defined as advertising that is tailored to an individual’s characteristics and/or interests or tastes (Hoy & Milne, 2010; Kelly et al., 2010; Sundar & Marathe, 2010). As consumers share a great deal of personal information (profile information, social relationships, interests, and behavior) on social network sites, marketers can use this information to personalize their advertising messages on social network sites to a great extent. This makes social network sites a very relevant context to study the effect of advertising personalization on customer responses.

61 Chapter 3: Is this for me?

Figure 3.1. Conceptual framework

Perceived relevance

Consumer Perceived response personalization (AB/CI)

Attitude toward Facebook

Research in other digital media shows that personalization can generate more favorable consumer responses because it increases the personal relevance of an ad (Anand & Shachar, 2009; Arora et al.,

2008; Iyer et al., 2005; Kalyanaraman & Sundar, 2006; Noar et al., 2009; Pavlou & Stewart, 2000).

Research on advertising personalization on social networking sites has become a growing area of research in recent years (e.g., Tucker, 2014; Walrave et al., 2016; Zarouali et al., 2018). Due to the specific context of social network sites, results from studies conducted in other digital environments may not apply here (Taylor, Lewin, & Strutton, 2011). First of all, Rodgers and Thorson (2000) state that, in order to understand how consumers react to marketing campaigns, it is necessary to understand why they use a medium. Users explicitly do not use SNSs for commercial goals (Ellison, Steinfield, &

Lampe, 2007). SNSs are primarily used to pass time and for amusement, next to relationship maintenance (Ku et al., 2013; Quan-Haase & Young, 2010). These motivations are different from e.g., e-mail or website visits in that these are primarily used for information seeking (Ku et al., 2013). E-mail can also be used for relationship maintenance, but to a far lesser extent than SNSs (Ku et al., 2013).

Furthermore, in a social network site, advertisements are displayed in an environment that is designed and controlled by the receiver of the message and is considered a personal space (Kelly et al., 2010).

It is, therefore, possible that users react negatively to personalized advertising on SNSs, because they may perceive personalization as disruptive or invasive, and hence more irrelevant than advertising in other online environments.

62 Chapter 3: Is this for me?

Because of the growing occurrence of personalization in this unique user-to-user ecosystem and the blurring lines between user and marketing content, the effects of personalized advertising through social networking sites warrant specific academic attention (Taylor et al., 2011). The first purpose of the current study is to investigate how advertising personalization impacts consumer responses (attitude toward the brand and click intention) to advertisements on social networking sites. The second purpose is to investigate the role of perceived ad relevance as a mediator between ad personalization on SNSs and consumer responses.

Even though the positive effects of personalization through perceived relevance have often been confirmed in prior (non-SNS) studies, the literature seems to lack an understanding of under what circumstances personalized ads can be (most) effective (Noar et al., 2009). More specifically, we will investigate the moderating role of the attitude towards the site on the relationship between ad personalization and consumer responses. The attitude toward the website has been shown to influence 3 consumer responses in other digital environments (Bruner & Kumar, 2000; Chen, Clifford, & Wells,

2002; Cho, 1999; Goldsmith & Lafferty, 2002; Stevenson et al., 2000). However, this study is the first to our knowledge to investigate its effect on personalized advertising messages. We will argue that a positive attitude toward the social network site reinforces the positive effects of personalized advertising.

We present the results of two experiments, set in a Facebook context in which we test the effects of ad personalization, perceived personal relevance, and the attitude toward Facebook on consumers’ brand attitude and click intention.

Literature Review

Advertising Personalization and Consumer Responses Advertising can be placed on a continuum ranging from no personalization, over rather general personalization or customization (e.g., sending local bridal shop ads to women whose relationship status is "engaged"), to full personalization (completely tailored or addressed to a particular individual based on his or her name, previous searches, web page visits, viewed content, or friends with connections to specific pages, groups, or applications) (Arora et al., 2008; Hawkins et al., 2008; Hoy & Milne, 2010).

Previous research in other digital environments has shown that personalization improves advertising effectiveness (Arora et al., 2008; Kalyanaraman & Sundar, 2006; Pavlou & Stewart, 2000; Tam & Ho,

63 Chapter 3: Is this for me?

2005). Two meta-analyses (Noar, Benac, & Harris, 2007; Sohl & Moyer, 2007) conclude that personalized messages are generally more effective than non-personalized messages in terms of being more memorable, more likable and sparking behavioral change. For example, personalization of a direct marketing message induces a better advertising response rate (Howard & Kerin, 2004). Research by

Abrahamse, Steg, Vlek, and Rothengatter (2007) showed that households exposed to a personalized message were more likely to adopt energy-saving behavior than participants in a control group.

Petty, Barden, and Wheeler (2002) build upon the Elaboration Likelihood Model (ELM) (Petty &

Cacioppo, 1986) to explain how tailoring or personalization can change attitudes and subsequent behavior. The ELM describes two routes to attitude formation in response to persuasive messages, the central and the peripheral route (Petty & Cacioppo, 1986). Consumers are likely to process via the central route when they are motivated and able to process this message. Peripheral processing, on the other hand, takes place when motivation and ability to process are low. In the logic of the ELM, personalization should benefit attitudes and behavior both under high and low elaboration (Petty et al.,

2002). Under conditions of high elaboration, perceived personalization can lead to biased message processing because personalized arguments could be perceived as stronger than non-personalized arguments. At the same time, under conditions of low elaboration, perceived personalization can serve as a heuristic cue that leads to a (albeit weak) positive attitude change.

The Mediating Role of Perceived Relevance

Several prior studies have examined the underlying mechanism of personalization effects (Kalyanaraman

& Sundar, 2006; Rimer & Kreuter, 2006; Tam & Ho, 2006). Researchers have proposed and tested several mediators for favorable personalization effects, such as self-referent thinking (Hawkins et al.,

2008; Tam & Ho, 2006), perceived uniqueness (Franke & Schreier, 2008) and feelings of accomplishment (Franke, Schreier, & Kaiser, 2010). Based on a review by Noar et al. (2009), increased personal relevance is identified as the primary mediator of positive personalization effects in many prior studies (Sundar & Marathe, 2010). This idea is also consistent with the Elaboration Likelihood Model

(ELM) (Petty et al., 2002; Petty & Briñol, 2010; Petty & Cacioppo, 1986).

Consumers tend to see a personalized message as more self-relevant because it uses information about themselves (Kalyanaraman & Sundar, 2006; Kim & Sundar, 2008; Tam & Ho, 2005; Zeng, Huang, &

64 Chapter 3: Is this for me?

Dou, 2009). This idea is in line with self-referencing, i.e., the extent to which a consumer relates information to oneself (Tam & Ho, 2006). Self-referencing can have a positive effect on attitudes under both central processing and peripheral processing (Hawkins et al., 2008; Romeo & Debevec, 1992). Under peripheral processing, self-referencing can be used as a decision aid (‘The product will be good because the advertisement is personalized for me.’)(Tam & Ho, 2006). At the same time, due to self- referencing, readers could also be more motivated to process personalized messages (and thus follow a more central route). When a message is perceived as more personally relevant, for example, because it is personalized, it does not only lead to greater attention, but also to greater elaboration, message processing, and ultimately, persuasion (Bright & Daugherty, 2012; Cho, 1999; Noar et al., 2009; Rimer & Kreuter, 2006; Tam & Ho, 2005). For example, Rimer and Kreuter (2006) argue that greater perceived relevance is the driver for personalized messages to generate more behavioral changes. Dijkstra (2005) found that enhancing standard smoking cessation materials with even a minimum amount of personal 3 information increased the perceived relevance of the communications (e.g., “directed at you personally”,

“takes into account your personal situation”) and the number of smoking quit attempts.

Especially in the context of online social networking sites, personal relevance seems a crucial condition in order for persuasion to occur in response to personalized messages, as one of the most important reasons for advertising avoidance in social network sites is a lack of perceived relevance (Kelly et al., 2010). As argued above, personalized advertising messages should be perceived as more relevant. Taylor et al. (2011) found that self-brand congruity and ad informativeness (related to relevance) positively influence users’ attitude toward social network advertising. Therefore, if advertising personalization manages to increase the perceived relevance of the advertisement, this should result in more positive consumer responses (Figure 3.1.).

A typical approach adopted by prior studies is to manipulate and compare the effect of two types of message, i.e., personalized versus non-personalized ones (Noar et al., 2009; Tucker, 2014). However, a number of studies have indicated that actual personalization (as manipulated by researchers) and perceived personalization (the degree to which a consumer perceives a match between a message and him-or herself) do not automatically match (e.g. Bettman, Luce, & Payne, 1998; Kramer et al., 2007;

Simonson, 2005). It is important that consumers perceive a message to be tailored to their needs and

65 Chapter 3: Is this for me? preferences before any favorable personalization effect can take place (Kramer, 2007). Consequently, perceived personalization is a more relevant construct than actual personalization (Kramer, 2007;

Kramer et al., 2007). Therefore, in the current research, we manipulate personalization by creating one personalized advertisement based on gender and one non-personalized advertisement, in order to induce variance in perceived personalization. However, as suggested by prior research, we use the measure of perceived personalization in our analyses. We therefore expect:

H1: Consumers who perceive an advertisement on a social network site as more personalized will have a more positive a) attitude toward the brand and a more positive intention to b) click the advertisement.

This effect is mediated by perceived relevance.

The Moderating Role of the Attitude toward the Social Network Site

Research indicates that positive personalization effects on consumer responses are influenced by moderating factors (see meta-analysis by Noar et al., 2007). For example, the effectiveness of e-mail personalization is likely to decrease with high personalization when consumers do not see a legitimate reason why their personal information is used (White et al., 2008). Also, highly personalized messages may not generate desirable responses from consumers who possess interdependent or collectivist tendencies (Kramer et al., 2007).

In the present paper, we propose that consumers’ attitude toward a social network site moderates their responses to personalized advertising on this site. Prior online research shows that a better attitude toward a website leads to a better brand attitude and purchase intention for embedded advertising

(Stevenson et al., 2000). Particularly for social networking sites, research shows that Facebook users who have a favorable attitude toward Facebook are more likely to purchase products that are advertised there (Wen et al., 2009). Lee and Ahn (2013) show that students who perceive Facebook as more trustworthy are also more likely to participate in a binge drinking prevention page on this site.

We propose that the positive effects of personalization and the attitude toward the social network site will reinforce each other (see Figure 3.1.). Consumers with a positive attitude toward the social network site may transfer this positive feeling toward embedded advertising. Fans of the social network site might find advertising on this site more informative or more entertaining in general, especially when the advertising is personalized. This should result in more positive consumer responses to the

66 Chapter 3: Is this for me? advertisement. Users who have a relatively more negative attitude toward the social network site may question the legitimacy of personalized advertising messages or the motives behind it.

H2: The positive effects of perceived personalization of advertisements on a social network site on the a) attitude toward the brand, and intention to b) click the advertisement is stronger for consumers with a more positive attitude toward the social network site than for consumers with a negative attitude toward this site.

Study 1

Experiment Design We conducted a between-subjects online experiment with two experimental conditions, in which we exposed a student sample to a fictitious Facebook home page. As mentioned, we manipulated personalization with one generic and one personalized condition based on gender, in order to induce 3 variance in perceived personalization. Chapter 2 of this dissertation found that gender was successful in eliciting perceived personalization. While Chapter 2 showed that there are elements relatively more important than gender to elicit personalization, we opted to test the effect of a moderately important personalization element. It is easy to conceive that extremely important personalization cues will benefit the perceived ad relevance, but we wanted to test whether a milder form of personalization would do the same. In line with the privacy calculus theory, stronger forms of personalization could also trigger adverse effects, which are not the focus of this chapter. The choice for gender is further inspired by the fact that it is commonly used in practice. In later chapters, we will test other personalization cues.

There are three types of advertising on Facebook (AdEspresso, 2016). The simplest one is a domain advertisement. This type of advertising is shown on the right-hand side of the screen, next to the user’s news feed. In the classification of the Interactive Advertising Bureau (2015), this type of advertising corresponds to a ‘medium rectangle’ display advertisement. A second type is the Page Post Link or

Newsfeed Ad, which includes a bigger image with text and a link description. In June 2014, a third format was released: the multi-product ad, through which an advertiser can promote several products and use up to three pictures. The present study tests the first type of advertisement.

The page was programmed in HTML to mimic an actual Facebook page and was filled with fictitious posts and activities that were identical for all participants (see Appendix 1.C.). To make the page feel

67 Chapter 3: Is this for me? more natural, respondents were also asked to provide their own name and the names of five friends, which were inserted into the page via piped text in the HTML code. On the right-hand side of the page was a medium rectangle display advertisement for a fictitious brand of perfume (Confiance), as can be found on a real Facebook page. We selected perfume because we thought this was a popular, affordable product with the target group. This product is also (increasingly) important in (online) sales

(Euromonitor International, 2015, July; Nielsen, 2014). A fictitious brand was used to avoid potential confounds of prior brand attitudes. Depending on the condition, the advertisement was either personalized (“For men with confidence” or “For women with confidence”, depending on the gender of the respondent) or generic (not personalized) (“Confiance”).

Participants

The study was conducted among students because they are among the heaviest users of Facebook

(Hoy & Milne, 2010). We recruited undergraduate and graduate students from a Belgian university via e-mail invitations containing a link to the online questionnaire. Participants (n = 155) ranged in age from 17 to 29 years old (M = 21.3, SD = 2.59) and 29.7% of the sample was male.

Procedure and measures

The questionnaire started with a welcome screen, with instructions on how to fill out the questionnaire.

Participants could proceed through the questionnaire at their own pace. They had to indicate their gender, age and education level and were asked to fill in their own name as well as the names of five friends before continuing to the experimental stimulus. Participants were randomly assigned to one of the two experimental conditions. Then, they were asked to complete the rest of the survey.

Participants rated the perceived degree of personalization of the ad (‘the information was fully tailored to my personal profile’). The mean and standard deviation for this measure indicate that there is indeed at least a certain degree of variance in the perceived personalization of the advertisements (푋̅ = 2.42,

SD = 1.51). We also measured the perceived relevance of the ad (2 items, e.g., ‘the information in the advertisement was relevant/useful’, α = .937) (Ahluwalia, Unnava, & Burnkrant, 2001), the attitude toward the advertised brand (Ab) (4 items, e.g., ‘I like the brand’, α = .787) (Lee, 2000), click intention for the ad (CI) (6 items, e.g., ‘It is likely that I will click this ad’, α = .941) (Chen et al., 2002), and participants’ attitude toward Facebook (4 items, e.g., ‘I think using Facebook is a good way to spent my

68 Chapter 3: Is this for me? time’, α =.735) (Chen & Wells, 1999; Ellison et al., 2007). All constructs were measured by means of

7-point Likert scales or semantic differentials. Construct scores were computed by calculating the average of the items per construct for use in further analyses.

Results To test our hypotheses, we analyzed the data using Hayes’ approach (2013) (model 5) with 1000 bootstrap samples (Figure 3.1.). We conducted two separate analyses for the two dependent variables, attitude toward the brand (Ab) and click intention (CI). In these two models, the perceived degree of personalization was entered as a continuous independent variable, the attitude toward Facebook as a moderator, and the perceived relevance of the advertisement as a mediator, all mean-centered.

Specifically, this model tests a) whether the indirect effect of perceived personalization on Ab and CI through perceived relevance is significant (H1), and b) whether the direct effect of perceived personalization on Ab and CI is significantly moderated by the attitude toward Facebook (H2). Due to 3 the highly feminine sample, we entered the gender of the respondent as a covariate in our analyses.

The results show a significant positive effect of perceived personalization on the mediator, perceived relevance (B = .586, p < .001) (Table 3.1.). In addition, perceived relevance has a positive and significant effect on Ab (B = .288, p < .001) and CI (B = .419, p < .001). Importantly, the indirect effect of perceived personalization through perceived relevance is positive and significant for both Ab

(.169, 95% CI = [.103; .258]), and CI (.246, 95% CI = [.155; .373]). These results support H11: the more a Facebook advertisement is perceived as personalized, the more relevant it is perceived, and the more positive the consumers’ attitude toward the brand and click intention. The results for Ab and CI suggest indirect-only mediation (Zhao, Lynch, & Chen, 2010): there was no direct effect of perceived personalization on Ab and CI when the mediator perceived relevance was added to the model. No significant interaction effect of perceived personalization and the attitude toward Facebook on Ab (B =

.023, p = .541) or CI (B = .051, p = .197) was found. H2 is rejected.

1 Note that the results from the analysis with actual personalization (corresponding to our manipulations, entered as a dummy variable) as the independent variable (see Appendix 3.A.). The results show that actual personalization does not have any significant effect on perceived relevance (B = -.127, p = . 578), nor on brand attitude (B = - .064, p = .653) or click intention (B = -.133, p = .345).

69 Chapter 3: Is this for me?

Table 3.1. Unstandardized Regression Weights (Study 1)

Perceived Ab CI relevance Perceived personalization .586*** .073 .023 Attitude toward Facebook - .026 .132* Perceived personalization * Attitude toward Facebook - .023 .051 Perceived relevance - .288*** .419*** Gender - .193 -.052 ;. R² .401 .285 .386 Note: *** p ≤ .001, ** p ≤ .010, * p ≤ .050

Discussion

The results of Study 1 confirm our expectation that perceived personalization positively affects consumer responses by increasing the perceived relevance of a Facebook ad. At the same time, our second hypothesis – a moderation effect of attitude toward Facebook – was rejected. The results for H1 suggest indirect-only mediation. This means that the direct effect of perceived personalization on brand attitude and click intention is no longer significant when accounting for perceived relevance. In terms of the

ELM, this would suggest central processing: perceived personalization works by increasing the motivation to process because the advertisement is considered more relevant. There is no direct effect of perceived personalization as a peripheral cue. The ELM predicts that people are more motivated to process an advertisement for a high involvement product than for a low involvement product (Franke,

Keinz, & Steger, 2009; Kalyanaraman & Sundar, 2006). A follow-up study (N = 60) indicates that perfume is indeed relatively high in involvement (M = 5.42, SD = .80). This might also explain why our second hypothesis was rejected. The attitude toward Facebook might serve as a more peripheral cue which does not exert any effect in case of a high involvement product. When processing a lower involvement product, on the other hand, consumers may be more likely to take the Facebook context into account as a peripheral cue (Noar et al., 2009; Petty & Briñol, 2010). Therefore, the effects of personalization and of the attitude toward the Facebook context may be different for high and low involvement products (Tam & Ho, 2005). That is why we conducted a second study, in order to test the robustness of the findings of H1 and to see whether H2 would be supported from a product with lower involvement. The follow-up Study indicated that vacuum cleaners are significantly lower in involvement

(M = 4.64, SD=.56) than perfume (t(26) = 2.98, p = .006).

70 Chapter 3: Is this for me?

Study 2

Experiment Design

A second online experiment was set up in order to test our hypotheses for a lower involvement product.

First, a pretest was conducted to select a low involvement product. Participants were randomly assigned to rate three products (out of a list of twelve) on the seven-point semantical differential scale of

Zaichkowsky (1985). We selected a vacuum cleaner as a low involvement product. The design and procedure of the second main study are identical to Study 1. We again used a fictitious brand, Dust

Devil. The advertising message was, again, either personalized (‘Because men know what is important’ or ‘Because women know what is important’ based on the gender of the respondent) or not personalized (“Dust Devil”).

Participants and Measures

Respondents were again undergraduate and graduate students recruited from a Belgian university, via 3 an e-mail containing a link to the questionnaire, and randomly assigned to one of the two conditions.

In total, 153 respondents with an age ranging between 18 and 30 (M = 21.7, SD = 2.56) participated in the experiment. 25.5% of the respondents were male. The measures are the same as in Study 1. All

Cronbach’s alphas were higher than .78. For this study, the mean for perceived personalization was

2.17 with a standard deviation of 1.49.

Results The same procedure as in Study 1 was used to test the hypotheses (Table 3.2.). The results again show a significant positive effect of perceived personalization on perceived relevance (B = .379, p < .001) and of perceived relevance on Ab (B = .254, p < .001) and CI (B = .316, p < .001). Perceived personalization also has a positive significant direct effect on CI (B = .206, p < .001), but not on Ab (B =.078, p = .211). Importantly, we again find a significant indirect effect of perceived personalization through perceived relevance for Ab (.096, 95% CI = [.047; .180]) and CI (.120, 95% CI = [.054; .223]), supporting H12. The results suggest indirect-only mediation for Ab and complementary mediation for CI

2 Note that the results show that actual personalization has a significant negative impact on perceived relevance (B = -.680, p = .002) and on brand attitude (B = -.327, p = .046), but not on click intention (B = -.083, p = .546; see Appendix 3.A).

71 Chapter 3: Is this for me?

(Zhao et al., 2010). Contrary to Study 1, we find a significant interaction effect of perceived personalization and the attitude toward Facebook on CI (B = .107, p = .003), although the effect on Ab is not significant (B = -.001, p = .971).

Table 3.2. Unstandardized Regression Weights (Study 2)

Perceived relevance Ab CI Perceived personalization .379*** .078 .206*** Attitude toward Facebook - -.003 .064 Perceived personalization * Attitude toward Facebook - -.001 .107** Perceived relevance - .254*** .316*** Gender - -.129 -.067

R² .176 .161 .438 Note: *** p ≤ .001, ** p ≤ .010, * p ≤ .050

We inspected the conditional direct effects of perceived personalization on CI for different values of participants’ attitude toward Facebook (Figure 3.2.), which are provided by the PROCESS macro (Hayes,

2013). We examine the effect of perceived personalization on CI separately for low, moderate and high values of the Attitude toward Facebook. The significant moderating effect on CI is found for both a moderate (β =.206, p < .001) and a positive (β = .345, p <.001) attitude toward Facebook, but not for a less positive attitude toward Facebook (β =.066, p = .412). Thus, the conditional direct effects for a moderate and a positive attitude toward Facebook show that there is a significant moderating effect for these levels of participants’ attitude toward Facebook on the relation between personalization and click intention. H2 is confirmed for CI, but not for Ab.

72 Chapter 3: Is this for me?

Figure 3.2. Conditional direct effect of perceived personalization on click intention (Study 2)

0.5

0.4 0.3 0.2

0.1

Click intentionClick 0

-0.1

-0.2 low (-1 SD) moderate high (+1 SD) Direct effect 0.066 0.206 0.345 LLCI -0.093 0.100 0.232 ULCI 0.226 0.311 0.457 Attitude toward Facebook 3 Notes: LLCI = lower limit confidence interval; ULCI = upper limit confidence interval Discussion

The results of Study 2 support the results of Study 1 that perceived personalization positively affects consumer responses by increasing the perceived relevance of a Facebook ad. In contrast with the results of Study 1, we do also find a residual direct effect of perceived personalization on click intention. This supports the idea that personalization also works as a peripheral cue (‘The product will be good because the advertisement is personalized for me.’) (Tam & Ho, 2006), especially in the context of lower involvement products.

Furthermore, contrary to Study 1, respondents’ attitude toward Facebook further moderated the effect of perceived personalization on click intention in this study. The positive effect of perceived personalization is only significant for individuals with a moderate to positive attitude toward Facebook. This finding indicates that the attitude toward Facebook serves as a peripheral cue, which will be relatively more important for a lower involvement product. In sum, while Study 2 supports the idea that advertising personalization on Facebook exerts a positive effect on ad responses, the study also suggests that different processing mechanisms are at the basis of these positive effects for relatively high and low involvement products.

73 Chapter 3: Is this for me?

General Discussion

The purpose of this study was to investigate the effect of perceived advertising personalization on a social network site, Facebook, on the attitude toward the advertised brand and click intention, taking into account the mediating role of perceived relevance and the moderating role of the attitude toward

Facebook. As such, the proposed moderated mediation model contributes to a better understanding of whether and when personalized ads can be more effective than non-personalized ads on a social network site.

In line with previous research in other contexts (Arora et al., 2008; Pavlou & Stewart, 2000; Tam & Ho,

2005), (perceived) personalization has a positive effect on consumer responses. It should be noted that the ads in our study used a very general personalization based on gender. Through Facebook, advertisers can personalize their advertisements to a much higher degree. It is therefore interesting to note that even this general form of personalization already induces positive effects, as long as it is perceived as personalized by consumers. This is consistent with the findings of Dijkstra (2005) and

Webb, Hendricks, and Brandon (2007) that even minimal degrees of personalization, or the mere prime of personalization, are sufficient to induce positive effects.

The results of the mediation analysis indicate that the effect of personalization on consumer responses occurs mostly indirect-only, through perceived relevance. As suggested in the literature (Petty et al.,

2002), personalized advertising seems to increase message processing by enhancing the perceived relevance of the message. This result signals that personalized advertising on social network sites can indeed be effective, but only insofar as they are indeed perceived by users as more relevant (Kramer,

2007).

The differences between Study 1 and Study 2 are consistent with differences in the processing of the advertisements due to involvement, based on the ELM. For higher involvement products, personalization seems to work more centrally, by increasing the perceived relevance of the advertisement. There is no remaining residual direct effect, nor is this effect moderated by the attitude toward Facebook. For lower involvement products, however, there seems to be a more peripheral effect at play as well. In Study 2, the direct effect of perceived personalization on click intention was significant, and the effect of personalization was further enhanced by a positive attitude toward Facebook. This finding suggests that

74 Chapter 3: Is this for me? users indeed use their attitude toward the website as a peripheral cue. While processing an advertisement for a high-involvement product such as perfume, the Facebook context might not be taken into account. However, for a low-involvement product (such as a vacuum cleaner in our study) consumers will incorporate peripheral cues, such as the Facebook context.

It might be that personalization fits a high involvement product better than it does low-involvement products. Indeed, a personally relevant message is most likely processed centrally when personal involvement is high (Noar et al., 2009). Taylor et al. (2012) found that product category involvement directly affects the self-enhancement value of a message. For a high involvement product, the motivation to process the advertisement is already quite high. positive effect of perceived personalization cannot be further increased by a positive attitude toward Facebook. This is also in support of prior findings that consumers who feel strongly about a product type are more likely to talk to others about that product or share the message (Taylor et al., 2012). 3

Implications Theoretical Contributions

The findings of the present study offer a number of theoretical contributions. The research partly fills the existing knowledge gap concerning advertising in a social network site context. This study experimentally tests how personalized advertising is processed and the circumstances in which the effect of personalization differs. Using the Elaboration Likelihood Model as a base, the proposed model was tested for two products differing in product category involvement. For both products a positive effect of personalized advertising on band attitude and click intention was found. This effect was stronger in

Study 1 with a higher involvement product than in Study 2, which is consistent with the ELM. Previous research also shows that a higher product involvement leads to more positive consumer responses for personalized advertising (Gordon, McKeage, & Fox, 1998; Kalyanaraman & Sundar, 2006).

The positive effects of perceived personalization can almost fully be attributed to the mediating role of personal relevance. If a personalized message is perceived as personally relevant, responses to the message are more positive. This result is in line with the ELM because personal relevance is believed to enhance motivation to process. This motivation is one of the key constructs in the ELM.

75 Chapter 3: Is this for me?

Furthermore, there was a difference in the processing mechanism for low and high involvement products. In case of a high involvement product, the attitude towards the SNS does not moderate the effect of personalization. However, for the low involvement product, the attitude toward Facebook can reinforce the effects of perceived personalization. The findings from this study suggest that the attitude toward Facebook can serve as a peripheral cue to enhance the effect of personalization on consumer responses.

The results of this study also help to shed light on the findings of previous studies (e.g. Maslowska,

Smit, & van den Putte, 2011) who did not find significant effects of personalization on consumers responses. This could be due to two reasons. First, a number of these studies used the manipulated actual personalization as a 0/1 variable in their analysis. Because actual and perceived personalization do not automatically match (Bettman et al., 1998; Nisbett & Wilson, 1977; Simonson, 2005), we used the perceived personalization as a basis for our analyses. Other authors have already argued that messages that are intended to be personalized can actually be interpreted by consumers as generic

(Kramer et al., 2007). Our findings indicate that perceived personalization has a significant, positive effect on consumer responses because it increases personal relevance. Using actual personalization as input might not be sufficient. We recommend future researchers to include perceived personalization in their analyses. Second, the results of this study indicate that for low involvement products, there is no significant effect of perceived personalization for users with a relatively negative attitude toward

Facebook. Therefore, studies that have tested personalized advertising without any context, or on a website which was not positively evaluated by respondents, will indeed have more difficulty to uncover positive effects of personalization. We recommend future researchers to either select positive contexts or at least measure the attitude toward the context.

Managerial Implications

Our results provide guidelines for practitioners. Personalized advertising on social networking sites leads to more positive consumer responses than non-personalized advertising. This is mainly because personalized ads are perceived as more relevant. When an ad is perceived as personally relevant, the attitude toward the brand and click intention will improve. It is especially important that consumers recognize personalized advertisements as personalized. If there is no accordance between the

76 Chapter 3: Is this for me? consumer’s characteristics, interests, etc., the advertisement will not be relevant, and consumer’s responses will not improve. This is more outspoken for a high-involvement product than for a low- involvement one. For low involvement products, perceived personalization also has a direct positive effect on click intention. Thus, in any case, brand managers should try to design their advertisements so that they will be perceived as personalized by their target audience. The results indicate that even a very general form of personalization, based on gender (which is something brand managers can easily apply) already induces positive effects, as long as consumers indeed recognize the advertisement as being more personalized. Thus, especially in the case of high-involvement products, advertisers should personalize their messages in such a way that they are perceived as particularly personally relevant. At the same time, brand managers should exert personalization efforts with care, because research indicates that when consumers are aware of personalization techniques, many consider this behavior as a violation of their privacy (Tucker, 2014; Turow, King, Hoofnagle, Bleakley, & Hennessy, 2009). 3

Brand managers should also consider which social networks to place their personalized advertising on.

SNSs that are better liked can reinforce the positive effect of perceived personalization. For a low involvement product, users with a moderate to positive attitude toward the site are more likely to click on a personalized ad. For a high involvement product, the attitude toward the social network site does not exert a significant effect. Therefore, it might be interesting, especially for advertisers of low - involvement products, to know the attitude toward the social network site they are advertising on and to select those sites that are well-liked by their target groups. Advertisers of high involvement products are less restricted by this limitation, and could therefore also opt for less popular sites, if advertising through these sites is cheaper, for example.

Finally, it is important to note that personalization is only one approach to increase personal relevance.

Separate analyses (for both studies) show that the direct effects of personal relevance on brand attitude

(Study 1: β = .512, p < .001; Study 2: β = .383, p < .001) and click intention (Study 1: β = .598, p

<.001; Study 2: β = .528, p < .001) are stronger than the direct effects of perceived personalization

(Study 1: βbrand attitude = .411, p < .001; βclick intention = .406, p < .001; Study 2: βbrand attitude = .256, p <

.001; except for click intention in Study 2: β = .535, p < .001). As such, it seems more important as advertiser to aim for relevance than for personalization. Personal relevance does not only have to do

77 Chapter 3: Is this for me? with the message being personalized based on personal characteristics, but timing with which the message is delivered, the frequency and the channels used are also meaningful to consumers (McMurtry,

2017). For example, when the consumer is not receptive for the message or when she has just bought the product that is being advertised than she will not perceive the message as being relevant, whether it is personalized or not.

Suggestions for future research

In the two experiments in the present study, we investigate products belonging to two opposite quadrants of the Rossiter and Percy Grid (Bergkvist & Rossiter, 2008; Rossiter & Percy, 1997). Percy and Rossiter (1992) have proposed a two-by-two matrix based on two dimensions: the level of perceived risk (product category involvement) and buying motivation (hedonic or utilitarian). Hedonic products

(e.g., designer clothes, sports cars, perfume, etc.) are used because of the aesthetic or sensory experience, for amusement, fantasy or fun. Utilitarian products (e.g., microwaves, personal computers, vacuum cleaners, etc.) are used for accomplishing functional or practical tasks (Wen et al., 2009).

Typically, hedonic products are evaluated on subjective characteristics, such as shape, taste or looks, while utilitarian products are more cognitively evaluated: consumers will focus on objective characteristics (Dhar & Wertenbroch, 2000). The advertised product in our first study (perfume) fits in the high involvement hedonic product category. The second product (vacuum cleaner) can be placed in the low involvement utilitarian category. It is important for further research to test products that represent other quadrants in the Rossiter-Percy Grid. The effect of involvement and hedonic-utilitarian products should, in other words, be disentangled. This can be especially important when one considers the motivations for using SNSs. SNS users primarily use SNSs to pass time and for amusement (Ku et al., 2013; Quan-Haase & Young, 2010), which is different from, for example, websites. As such, hedonic product advertisements may be more congruent with consumers’ motives (and thus perhaps more easily perceived as self-relevant) than ads for utilitarian products, because SNS users are not looking for information.

Second, the current study was conducted with a student sample. Even though this is a highly relevant sample to study advertising effects in social network sites, and especially Facebook (Hoy & Milne, 2010), the robustness of the model could be tested in different samples as well. Moreover, the sample was

78 Chapter 3: Is this for me? highly female, which might potentially affect the findings. Even though gender was entered as a covariate to control for the direct effect of gender, it did not control for the moderating effects. Future research should examine gender and other socio-demographic characteristics as a moderator.

Future research should also examine other potential moderators of personalization effects on social network sites. For example, personality traits - such as the need for uniqueness (Maslowska, van den

Putte, & Smit, 2011), privacy concerns (Maslowska, van den Putte, et al., 2011; Pfiffelmann et al., 2019; Sundar & Marathe, 2010; Yu & Cude, 2009), need for cognition (Tam & Ho, 2005), self-referent thinking (Hawkins et al., 2008; Tam & Ho, 2006), extraversion and conscientiousness (Chiu, Hsieh, Kao, & Lee,

2007), and the attitude toward personalization could be investigated. Maslowska, van den Putte, et al.

(2011) show that consumers high in need for uniqueness were more negative toward a generic newsletter compared to a personalized newsletter. This is due to the fact that consumers high in need for uniqueness derive satisfaction for the perception they are unique (Simson & Nowlis, 2000) and are 3 better in memorizing information that distinguishes them for others (Codol, Jarymowicz, Kaminska-

Feldman, & Szuster-Zbrojewicz, 1989). Moreover, we measured attitude toward the social networking site by using items from Chen and Wells (1999) and Ellison et al. (2007). However, attitudes toward the social networking site could consist out of different aspects (e.g. attractiveness, sense of community, level of entertainment, security, etc.). As a result, future research could examine which of these aspects is driving the moderating role of attitude toward the social networking site.

The present study includes attitude toward the brand and click intention as the dependent variables.

While models like the Theory of Reasoned Action and the Theory of Planned Behavior (e.g. Ajzen, 2002) posit that attitudes and behavioral intentions are important antecedents of actual behavior, and this is confirmed in many studies (e.g., Armitage & Conner, 2001; De Cannière, De Pelsmacker, & Geuens,

2009), the relationship between intention and behavior is not perfect (e.g. Sheeran, 2002; Van Ittersum,

2012). Future research should consider including measures of actual behavior, such as clicking, purchasing or ad forwarding or other word-of-mouth behavior.

Because of technical limitations, in the present study participants were exposed to a static Facebook page. Although the Facebook pages were adjusted to mimic a real Facebook home page, the lack of

79 Chapter 3: Is this for me? interactivity could have influenced the results. The external validity of the experiment could be improved by using an interactive or even a real social network environment.

We used only one type of personalization, based on gender. This means that our ad was only slightly personalized. Even though results by Dijkstra (2005) and Webb et al. (2007) suggest that even minimal degrees of personalization, or the mere prime of personalization, are sufficient to induce positive effects, effects could have been stronger with more “extreme” forms of personalization. For example, degrees of personalization can be placed on a bipolar linear continuum framework between “no personalization” and “full personalization” (completely tailored or addressed to a particular individual) (e.g., Arora et al.

2008; Hawkins et al. 2008). Future research could examine these different degrees of personalization to test whether they would have an impact on consumer responses in a different way. It is likely that higher degrees of personalization would lead to negative responses, especially in SNSs, because advertisements are displayed in an environment that is considered a personal space (Kelly et al., 2010).

Research on web care (van Noort & Willemsen, 2012), for example, indicates that consumers value proactive brand communications on brand-generated, but not consumer-generated platforms. It is, therefore, possible that users react more negatively to higher degrees of personalized advertising on

SNSs, because they may perceive personalization as disruptive or invasive, and hence more irrelevant than advertising in other online environments. It would, therefore, be interesting to more explicitly compare consumer responses to personalized advertising on SNSs to other online environments

How and When Personalized Advertising Leads to Brand Attitude, Click and WOM Intention1

1 An earlier version of this chapter was presented at 2018 International Conference of Research in Advertising, Valencia, Spain. 4 Chapter 4: How and When

Abstract

We study the effect of perceived personalization in advertising on social networking sites (SNS) on consumer brand responses. In Study 1 (N=202), we test a parallel mediation via perceived personal relevance and intrusiveness on brand attitude (Ab) and click intention (CI). Perceived personalization improves Ab and CI by increasing the perceived personal relevance and, unexpectedly, by decreasing the perceived intrusiveness of the ad. A second experiment (N = 264) confirms these findings and extends them to positive word-of-mouth (WOM) intention. Study 2 also examines the moderating role of the attitude toward the SNS and of perceived privacy protection by the SNS. Both the attitude toward the SNS and perceived privacy protection by the SNS moderate the indirect effect of perceived personalization via personal relevance on click intention, but not on Ab or positive WOM intention. More specifically, more positive attitudes toward the SNS and higher levels of perceived privacy protection by the SNS strengthen the positive indirect effect of perceived personalization on CI through perceived relevance. Study 2 further extends the processing mechanism of personalized advertising by additionally including the mediating effects of perceived entertainment, reactance, and self-brand connection.

Perceived personalization has a positive indirect effect on self-brand connection via perceived personal relevance and via perceived entertainment, but not via perceived intrusiveness. Self-brand connection, in turn, has a positive effect on consumers’ responses. Perceived personalization also has a negative indirect effect on reactance to the advertisement via perceived relevance, entertainment, and intrusiveness. Contrary to expectations, however, reactance does not significantly affect brand responses.

82 Chapter 4: How and When

Introduction

Social media have taken on a central role in many people’s lives: in April 2019, there were 3.48 billion social media users, which was an increase of 9% compared to 2018 (Kemp, 2019). The popularity of social media makes them attractive platforms for advertisers. For example, Facebook, the most popular social networking site, reported more than 50 billion dollars in advertising revenue in 2018 (Facebook

Inc., 2019c). In an attempt to make advertising more effective, social media platforms provide opportunities to tailor or personalize advertising to users based on their user profiles (Facebook, 2008).

Advertising personalization can improve advertising effectiveness as it increases the perceived personal relevance of an ad (De Keyzer et al., 2015; Walrave et al., 2016). At the same time, personalized advertising can also lead to feelings of intrusiveness, which is detrimental for advertising effectiveness (Pfiffelmann et al., 2019). Although prior research documents these two processing mechanisms, there are no studies, to our knowledge, that have empirically tested them as competing hypotheses in a single study. Our first contribution is that we test the mediating effects of perceived personal relevance and 4 intrusiveness simultaneously, to assess their relative importance in explaining the effect of advertising personalization on brand attitude, click intention (Study 1 and 2) and positive word-of-mouth (WOM) intention (Study 2).

Personalized advertising may not always be more effective than generic advertising, and research indicates that favorable personalization effects are subject to moderation by external factors (Noar et al., 2007; Pfiffelmann et al., 2019). For example, the effectiveness of personalized advertising decreases when consumers do not see a legitimate reason why their personal information is used (Pfiffelmann et al., 2019; White et al., 2008). The second contribution of our study is that we investigate the moderating role of two perceptions relating to the social networking site. The first moderator is users’ attitude toward the social networking site. Previous research shows that people’s attitude toward the social networking site affects the direct relationship between personalized advertising and click intention (De

Keyzer et al., 2015). However, their moderation pertains to the direct effect only, and not to any indirect effects. The current study, therefore, adds to a deeper understanding by testing moderated mediation.

The second moderator under study is SNS users’ perceived privacy protection by the SNS. (Malhotra et al., 2004) suggest that consumers who do not feel informed about a company’s privacy policies will be

83 Chapter 4: How and When reluctant to provide personal information. They may, therefore, also respond more negatively to personalized advertising using that information.

The third contribution of the current research lies in extending our understanding of the processing mechanism of personalized advertising. We extend our initial model containing perceived personal relevance and intrusiveness (Study 1) by adding perceived entertainment - the extent to which an advertisement provides pleasure, diversion or amusement (Taylor et al., 2012) - as an additional first- order parallel mediator (Study 2). Perceived entertainment is a more affective response, where perceived personal relevance and intrusiveness make up more evaluative responses. Furthermore, self- brand connection and reactance to the advertisement are added in parallel as second-order serial mediators (Study 2). Self-brand connection is the connection consumers establish between their self- concept and the identified brand meaning and its benefits (Liu & Mattila, 2017). This connection results in more positive responses toward the brand which can be explained by the fact that the brand addresses their psychological needs (Escalas & Bettman, 2003; Palazon, Delgado-Ballester, & Sicilia,

2018), such as receiving relevant information or being entertained without intrusions. Reactance to the advertisement, on the other hand, refers to a (negative) psychological response consumers might experience in response to advertisements they consider as inappropriate or intrusive and in which they perceive their freedom as being threatened (Miron & Brehm, 2006; White et al., 2008; Young & Kim,

2019). Reactance causes consumers to ignore the ad or to developing negative responses (van Doorn

& Hoekstra, 2013; White et al., 2008; Young & Kim, 2019). By adding these mediators, which have largely been overlooked in the literature on advertising personalization, the paper contributes to a more exhaustive understanding of both the positive and the adverse personalization effects found in previous research.

Through these two studies, set in a Facebook context, we offer a deeper theoretical understanding of how personalized advertising affects brand responses through personal relevance, intrusiveness, entertainment, self-brand connection, and reactance. Additionally, we explore the boundary conditions of the process by studying the role of the attitude toward the social networking site and perceived privacy protection by the social networking site. Advertising practitioners may use the insights of our

84 Chapter 4: How and When two studies to enhance their brand attitude and stimulate consumer brand engagement by personalizing advertisements.

Study 1 Literature Review and Hypotheses Personalized advertising is advertising tailored to an individual’s characteristics and/or interests or tastes, for example, based on information from social media (Walrave et al., 2016). Previous research shows that personalization can improve advertising effectiveness as consumers generally perceive advertising that uses information about themselves as more personally relevant (e.g., De Keyzer et al.,

2015; Tucker, 2014; Walrave et al., 2016; Zeng et al., 2009). For example, a sixteen-year-old girl is likely to find an ad for a concert by her favorite artist in her home town more relevant than an ad for incontinence products for men sold in another country. The extent to which a consumer relates information to himself or herself (self-referencing) positively influences attitudes (Hawkins et al., 2008).

Self-referencing serves as a decision aid or a heuristic cue (Tam & Ho, 2006) or could motivate readers 4 to process the personalized message further and thus lead to more considerable attention, elaboration, message processing and ultimately persuasion (Bright & Daugherty, 2012; Tam & Ho, 2005). According to Kelly et al. (2010), a lack of perceived personal relevance is the leading cause of advertising avoidance in social networking sites (SNSs). Relevance thus seems an essential condition for persuasion in online SNSs. Therefore, we hypothesize:

H1: Consumers who perceive an advertisement on a social networking site as more personalized will have a more positive a) brand attitude and b) click intention, via the mediating effect of perceived personal relevance of the advertisement.

Although personalized advertising may have a positive impact on consumer responses by increasing its perceived personal relevance, higher levels of personalization might also be perceived as intrusive (Li,

Edwards, & Lee, 2002), which could prevent consumers from processing the ad (Morimoto & Chang,

2006). Intrusiveness is “a psychological reaction to ads that interferes with a consumer’s ongoing cognitive processing” (Li et al., 2002, p. 39). Pfiffelmann et al. (2019) find that advertising that uses more personal data increases feelings of intrusiveness, which negatively affects purchase intentions.

85 Chapter 4: How and When

We thus expect a negative effect of personalization on brand responses, via an increase in perceived intrusiveness:

H2: Consumers who perceive an advertisement on a social networking site as more personalized will have a more negative a) brand attitude and b) click intention via the mediating effect of perceived intrusiveness of the advertisement.

Both mediating effects have been found separately in prior research. However, studies related to personalized advertising in social media have failed to address the potential trade-off between the increase in perceived personal relevance and perceived intrusiveness triggered by personalization (van

Doorn & Hoekstra, 2013). The privacy calculus model suggests that the extent to which people exercise privacy practices is based on a cost-benefit trade-off (Chen, 2018). While privacy calculus is mainly used to explain people’s behavior in disclosing information online or other privacy protecting behaviors (e.g.,

Dienlin & Metzger, 2016), parallels can be drawn to their responses to personalized advertising

(Demmers et al., 2018). Receiving more relevant content and advertisements is one benefit of disclosing information. At the same time, the intrusiveness of ads using personal data, could be considered a cost.

In determining their overall evaluation of a personalized ad, people will consider whether the benefits outweigh the costs. Some studies suggest that users are becoming more and more concerned about privacy on the Internet and particularly about how their data are used to personalize advertisements

(Aguirre et al., 2015; Chen, 2018). At the same time, a body of research has documented the phenomenon of privacy paradox, where people continue to reveal their personal information in exchange for benefits, in spite of reported privacy concerns. This phenomenon is particularly prominent on social media (Chen, 2018; Xie & Karan, 2019). Although the privacy calculus describes the potential trade-off between benefits and costs but does not prescribe whether one effect should be greater than the other. It is therefore unclear what to expect of the relative strength of the two mediating effects.

Figure 4.1 shows the conceptual model of Study 1.

RQ1: What is the relative strength of the indirect effects ad personalization on a) brand attitude and b) click intention through perceived personal relevance and perceived intrusiveness of the advertisement?

86 Chapter 4: How and When

Figure 4.1. Conceptual framework Study 1

Study Design and Procedure

To test our conceptual framework (Figure 4.1.), we set up a between-subjects experiment with four conditions (control condition, age-based personalization, gender-based personalization, interest-based personalization; see Appendix 1.D. for the advertisements). We focused on Facebook because it is the 4 most popular social networking site today in term of users (Statista, 2019). We created a mock Facebook newsfeed containing a mock profile picture update (a nature scene) as filler, followed by a native advertisement for a fictitious travel agency (‘Lovely Travel’). The newsfeed and advertisement were developed to resemble the real Facebook environment to enhance the external validity of the study.

The organization was fictitious to avoid possible confounds due to prior experience. Personalization was manipulated through the slogan. The ads were otherwise identical across conditions. In the non- personalized control condition, the slogan was generic: ‘Lovely Travel, discover why others were enchanted by this travel agency!’. In the personalized conditions, either participants’ gender (‘Lovely Travel, discover why other [women/ men] were enchanted by this travel agency!’), age (‘Lovely Travel, discover why others in their [twenties/ thirties/ forties/ fifties/ sixties] were enchanted by this travel agency!’) or interests (‘Lovely Travel, discover why others who are interested in [cycling/ nature/ walking/ going out/ skiing/ wellness/ shopping/ culture] were enchanted by this travel agency!’) were inserted in the slogan based on the information they provided at the start of the questionnaire (see below). The different personalization elements were used to create variance in the perceived

87 Chapter 4: How and When personalization scores. Chapter 2 indicates that gender, age, and interests are all useful personalization elements to elicit perceived personalization.

Participants (N = 202, 62.9% between 20 and 29 years, 29.7% male) were recruited via a Facebook snowball sample. Participants were randomly assigned to one of the four experimental conditions and could proceed through the questionnaire at their own pace. The questionnaire started with a welcome screen in which participants’ anonymity was ensured. Participants had to indicate their gender, age, and interests from a list (cycling, nature, walking, going out, skiing, wellness, shopping, and culture). This information was used to manipulate the personalization of the advertisements. Participants then reported how much time they spent on Facebook on a regular day. Most spent less than one hour

(82.3%) on Facebook on a regular day. They were instructed to attentively look at the mock Facebook newsfeed (containing one of the four advertisements) as if it was their own. Next, participants rated the perceived degree of personalization, adapted from Srinivasan et al. (2002) and Kalyanaraman and

Sundar (2006), perceived personal relevance, from De Keyzer et al. (2015), and perceived intrusiveness, from Li et al. (2002), of the advertisement on five-point Likert scales or bipolar semantic differentials.

Finally, participants rated their brand attitude on a five-point bipolar semantic differential from

Srinivasan et al. (2002). Table 4.1. provides an overview of the different measures.

Table 4.1. Measures (Study 1)

Measure Items Cronbach’s Mean (SD) alpha Perceived personalization This ad is tailored to my situation. .889 3.130 I believe this ad is customized to my needs. (.879) This ad was targeted at me as a unique individual. I believe that this ad is customized to my characteristics. This ad was personalized according to my profile. Perceived personal What do you think about the advertisement? .866 2.746 relevance Not important – Important (.841) Not relevant – Relevant Meaningless – Meaningful Perceived intrusiveness How does this ad appear to you? .740 2.896 Not distracting – Distracting (.799) Not disturbing – Disturbing Not forced – Forced Not interfering – Interfering Not intrusive – Intrusive Brand attitude In general, what is your feeling toward travel .786 2.860 agency Lovely Travel? (.607) I don’t like it – I like it. The agency doesn’t seem reliable. – The agency seems very reliable.

88 Chapter 4: How and When

The agency is of very low quality. - The agency is of very high quality.

Time spent on Facebook In the past week, on average, how many minutes per day have you spent on Facebook? Not daily. Less than 10 minutes. Between 11 and 30 minutes Between 31 minutes and 1 hour Between 1 and 2 hours Between 2 and 3 hours More than 3 hours Results

We used Hayes’ (2013) PROCESS macro (Model 4) with 5,000 bootstrap samples to test hypotheses 1 and 2 and answer the research question (see Table 4.2.). Based on previous research (e.g., Kramer,

2007; Li, 2016), we argue that perceived personalization is a more relevant construct. Moreover, using perceived personalization as the independent variable allows to generalize the findings beyond the specific personalization cues used in the current study. Thus, perceived personalization was the independent variable. Perceived personal relevance and perceived intrusiveness were entered as mediating variables. We ran two distinct analyses, one with brand attitude as the dependent variable 4 and one with the click intention as the dependent variable. We report the completely standardized indirect effects (for a complete discussion, see Hayes, 2018).

Table 4.2. Unstandardized regression weights with perceived personal relevance and perceived intrusiveness as mediators (Study 1) Perceived personal relevance Perceived intrusiveness Ab B Sig b Sig b Sig Constant 1.526 <.001 3.755 <.001 2.614 <.001 Perceived personalization .465 <.001 -.183 .009 -.001 .991 Perceived personal relevance .279 <.001 Perceived intrusiveness -.187 .002 Gender -.019 .876 -.133 .299 -.052 .552 Age -.096 .104 -.053 .489 .041 .272 R² .253 .048 .249

The indirect effects of perceived personalization on brand attitude (b = .187, CI [.100; .285] and click intention (b = .175, CI [.100; .270] via perceived personal relevance are positive and significant, confirming H1. H2 is not supported because the indirect effects of perceived personalization on brand attitude (b = .049, CI [.009; .101]) and click intention (b = .037, CI [.006; .080] via perceived intrusiveness are positive and significant, which is in the opposite direction of the hypothesis. The pairwise comparisons of the indirect effects show that the indirect effect on both brand attitude (brelevance

89 Chapter 4: How and When

– bintrusiveness = .138, [CI: .036; .246]) and click intention (b = .138, CI [.055; .232] via perceived personal relevance is significantly stronger than the indirect effect through perceived intrusiveness. This answers our research question1.

Discussion

Study 1 confirms earlier findings on the importance of perceived personal relevance (e.g., De Keyzer et al., 2015; Kalyanaraman & Sundar, 2006). With increasing levels of perceived personalization, perceived personal relevance also increases, leading to a more positive brand attitude and click intention. In line with Young and Kim (2019), we expected that using perceived personalization would also increase the perceived intrusiveness of the ad. However, our results indicate the opposite: the more an advertisement is perceived as personalized, the less it is experienced as intrusive. A possible explanation may be that personalized advertisements offer more valuable information than non-personalized advertisements and may, therefore, be less unwanted. Other more recent research also fails to document negative effects of personalization. Kim and Han (2014) found that the personalization of mobile advertisements is negatively related to perceived irritation (a concept related to intrusiveness): Higher personalization levels correspond to lower perceived irritation. Pfiffelmann et al. (2019) found that personalization did not exert a direct effect on perceived ad intrusiveness (although they found an indirect effect through visual attention). At the same time, the increase in visual attention to a personalized advertisement also helped to reduce (negative) attitudinal persuasion knowledge.

The level of perceived personalization in Study 1 is also modest (M = 3.130). It is possible that more extreme perceptions of personalization would produce the anticipated increased in intrusiveness. We, therefore, set up a second study to test the robustness of the findings from Study 1 with more diverse manipulations of personalization. As suggested in Chapter 2, combining personalization elements in a single advertisement would induce stronger perceptions of personalization.

1 Note that the results from the analysis with actual personalization (corresponding to our manipulations, entered as a dummy variable) as the independent variable (see Appendix 3.B.). The results from the analysis with the actual personalized conditions as the independent variables (the control condition is used as reference category). The results show that actual personalization does not significantly affect perceived relevance (Bgender = -.162, p = .349; Bage = -.042, p = .807; Binterests = -.118, p = .483). It does affect perceived intrusiveness (Bgender = .490, p = .003; Bage = .306, p = .091; Binterests = .363, p = .034).

90 Chapter 4: How and When

The second aim of Study 2 is to test the boundary conditions of the mediating effects of relevance and intrusiveness. We propose that consumers’ attitude towards the social networking site and the perceived privacy protection by the social networking site moderate responses to personalized advertising. Previous research in other digital environments indicates that people’s attitude toward the website influences their ad responses (Chen et al., 2002; Goldsmith & Lafferty, 2002). In Chapter 3, we showed that people’s attitude toward the social networking site moderates the direct relationship between perceived personalization and click intention. In this chapter, we investigate the moderation of the indirect effect. Theoretically, the latter is more interesting. Our second study, therefore, adds to a deeper understanding by testing moderated mediation. A second moderator is the perceived privacy protection by the social networking site. Consumers’ concerns over their privacy protection determines ad skepticism and ad avoidance (Baek & Morimoto, 2012), but also the likelihood to click on an ad (Tucker, 2014). Study 1 is based on a fairly young sample. The experience with and the appreciation of personalized advertising may differ between young adults and older age groups. Young adults are typically heavy users of social media (West, 2019). Van den Broeck, Poels, and Walrave (2015) found 4 that older respondents express more privacy concerns than younger people. Yet they reported modifying privacy settings less frequently than the younger age groups. In order to test the moderating effects of the attitude toward the social networking site and people’s perceived privacy protection by the social networking site, Study 2 recruits a more representative sample of social media users.

Finally, we extend the processing mechanism through which perceived personalization operates by adding three additional mediating variables, namely perceived entertainment, self-brand connection, and reactance to the advertisement (see Figure 4.2.). Study 2 also adds positive word-of-mouth as another behavior-oriented dependent variable, next to click intention (and brand attitude). Positive word-of-mouth intention represents viral marketing intentions (Chang, Yu, & Lu, 2015). Click intention is vital because in online advertising in general, and SNS advertising in particular, pay-per-click (PPC) billing has become standard practice (Asdemir, Kumar, & Jacob, 2012).

91 Chapter 4: How and When

Study 2

Investigating Boundaries

Figure 4.2. Conceptual framework Study 2 (Investigating Boundaries)

The Moderating Role of the Attitude toward the Social Networking Site

Consumers with a more positive attitude toward the social networking site respond more positively to advertising on the social networking site in general, and personalized advertising in particular (De Keyzer et al., 2015). In this chapter, we try to further break down the mechanism through which a positive attitude toward the social networking site reinforces the positive effect of personalization on brand responses. First, we focus on the path through perceived relevance. Facebook users who have a favorable attitude towards Facebook are more likely to purchase products advertised (Wen et al., 2009).

Their favorable predisposition towards the site could make users feel that personalized messages are even more personally relevant to them as they may interpret these as a genuine attempt of the site to provide them with relevant information adapted to their personal needs. Positive perceptions of ad relevance may also affect brand responses more positively if users with a more positive attitude toward the social networking site reward the advertised brand for the effort they put into tailoring their ad content.

Second, a positive attitude toward the social networking site might also reinforce the indirect effect of perceived personalization on consumer responses via perceived intrusiveness. Consumers with a positive attitude toward the social networking site might find personalized advertisements less intrusive because

92 Chapter 4: How and When the positive attitude carries over to their advertising perceptions. White et al. (2008) found that the effectiveness of personalized advertising decreases when consumers don’t see a legitimate reason for the use of their personal information. A positive attitude toward the social networking site may mitigate reservations about the legitimacy of personalized advertising and would, therefore, lead to a (further) reduction in its perceived intrusiveness. At the same time, users with a more positive attitude toward the social networking site may be more forgiving of a single transgression and brand responses would not suffer as much from perceived intrusiveness. Therefore, we expect:

H3: The positive effect of perceived personalization via perceived personal relevance on a) brand attitude, b) click intention and c) positive word-of-mouth is stronger when users’ attitude toward the social networking site is more positive.

H4: The positive effect of perceived personalization via perceived intrusiveness on a) brand attitude, b) click intention and c) positive word-of-mouth is stronger when users’ attitude toward the social networking site is more positive. 4

The Moderating Role of Perceived Privacy Protection by the Social Networking Site

Actions creating trust, such as privacy protection measures and transparency, can lead to a higher acceptance of online advertising personalization (Li & Unger, 2012). Aguirre et al. (2015) found that consumers exhibit higher click intentions for a highly personalized advertisement on a trusted website compared to an untrustworthy website. In a field experiment, Tucker (2014) showed that enhancing users’ perceived control over their privacy makes them nearly twice as likely to click on personalized ads. Perceived privacy control has a negative influence on privacy concerns and, indirectly, on their perceptions about the ad (Mpinganjira & Maduku, 2019). When users are confident that their privacy is protected, they may appreciate personalized ads more. This appreciation could result in a boost of the perceived relevance of personalized advertising. They may also value the perceived relevance more profoundly, with a resulting synergy effect on brand responses.

At the same time, negative perceptions of privacy protection should aggravate the perceived intrusiveness of personalized advertising. Consumers are more willing to be profiled for personalized advertising when they perceive a company’s privacy management policies as effective (Awad &

Krishnan, 2006). They are likely to feel more reassured that they could explicitly prevent the use of

93 Chapter 4: How and When unwarranted information and would, therefore, experience the ads as less intrusive. Consumers lacking knowledge about a company’s privacy policies will already be reluctant to provide more personal information (Li & Unger, 2012; Malhotra et al., 2004). When this information is then used in a personalized advertisement, this is deeply disturbing. Tucker (2014) found that personalization was relatively more effective for personalized ads that used unusual information after perceptions of privacy control were enhanced. They speculate that users were concerned that the information used was simply too personal to be exploited in advertising without a corresponding sense of control over their data.

Mpinganjira and Maduku (2019) further argue that, by enhancing trust, privacy policies can negate the negative consequences of perceived ad intrusiveness on the perceived ethical value of a brand (and therefore, on brand attitude and intention). Therefore, we expect:

H5: The positive effect of perceived personalization via perceived personal relevance on a) brand attitude, b) positive word-of-mouth and c) click intention is stronger when users’ perceived privacy protection by the social networking site is higher.

H6: The positive effect perceived personalization via perceived intrusiveness on a) brand attitude, b) positive word-of-mouth and c) click intention is stronger when users’ perceived privacy protection by the social networking site is higher.

Extending the Mechanism

The final contribution of this paper lies in the extension of the processing mechanism of personalized advertising. The first extension is the incorporation of perceived entertainment - the extent to which an advertisement is perceived as fun and entertaining (Xu, Oh, & Teo, 2009) - as a more affective first- order mediator, where relevance and intrusiveness are more evaluative (Figure 4.3.). Second, we add two second-order mediators to the processing mechanism. The first second-order mediator is self-brand connection, which refers to the connection consumers establish between themselves and the brand (Liu

& Mattila, 2017). This mediator can add to the explanation of positive personalization effects found in previous research. The second second-order mediator is reactance to the advertisement, which refers to the psychological reactance that can occur when encountering an advertisement (White et al., 2008).

This mediator can add to the explanation of adverse personalization effects previously found (van Doorn

& Hoekstra, 2013; White et al., 2008).

94 Chapter 4: How and When

Figure 4.3. Conceptual framework Study 2: Extending the Mechanism

The Mediating Role of Self-Brand Connection

For each consumer brands consist of a complex set of associations that reflects the brands’ attributes and benefits (Okazaki & Taylor, 2013). Self-brand connection refers to the match between consumers’ self-concept (actual and ideal self) and the brand associations that consumer has (Liu & Mattila, 2017). The process of self-brand connection follows two steps. First, the advertising message is processed and the brand meaning, and benefits are identified. Next, consumers try to match their self-concept (both the actual self and ideal self) with the identified brand meaning and benefits (Liu & Mattila, 2017). 4 Escalas (2004) argues that a brand might be used to fulfill psychological needs. For example, owning Apple products provides a sense of creativity. Consumers who identify themselves as creative (or who would like others to see them as creative) are more likely to form a self-brand connection with Apple than those who are less creative. Because personalized advertising addresses the consumer based on his/her needs and interests, the advertising message and the embedded brand will more likely be perceived as matching their own self-concept.

Perceived personal relevance. Perceived personalization might positively affect self-brand connection as a result of perceived relevance. The findings of Study 1 indicate that perceived personalization increases relevance. When being confronted with personalized advertising, a consumer will try to relate the information to himself or herself (self-referencing) (Hawkins et al., 2008). This self- referencing will increase the likelihood that a consumer can match the brand to his/her self-concept and therefore increase the likelihood to build a self-brand connection. In sum, we expect that perceived personalization will increase self-brand connection via perceived personal relevance.

95 Chapter 4: How and When

H7: Consumers who perceive an advertisement on a social networking site as more personalized will have a stronger self-brand connection via the mediating role of perceived personal relevance.

Perceived entertainment. Perceived personalization might also positively affect self-brand connection via perceived entertainment. Perceived entertainment reflects the extent to which consumers perceive an advertisement as providing pleasure, diversion, or amusement (Taylor et al., 2012).

Consumers find personalized mobile advertisements more entertaining (Kim & Han, 2014). What consumers perceive as entertaining is highly personal and reflects who a person is and how they see themselves (Taylor et al., 2012). For example, when a person sees himself as tech-savvy he might consider advertisements for the latest tablets entertaining because they show the latest features and keep him up to date about the latest tech products. Therefore, a personalized advertisement that addresses a consumer personally might be perceived as more entertaining. Previous research indicates that perceived entertainment is an important determinant of consumers’ responses to advertising (e.g.,

Kim & Han, 2014; Taylor et al., 2011). When the advertisement is perceived as entertaining, this might carry over to the brand and therefore establish the idea that brand is entertaining. In a second step, a consumer might try to match this idea with their self-concept and create a self-brand connection.

Therefore, we expect:

H8: Consumers who perceive an advertisement on a social networking site as more personalized will have a stronger self-brand connection via the mediating role of perceived entertainment.

Perceived intrusiveness. A third path via which perceived personalization might affect self-brand connection is via perceived intrusiveness. The findings of Study 1 already suggested that perceived personalization decreases perceived intrusiveness. Nevertheless, when an advertisement is intrusive, this might hinder consumers to relate their self-concept to that of the brand, impeding the development of self-brand connection. Therefore, we expect:

H9: Consumers who perceive an advertisement on a social networking site as more personalized will have a stronger self-brand connection via the mediating role of perceived intrusiveness.

Furthermore, self-brand connection should benefit brand responses (De Pelsmacker, Geuens, & Van

Den Bergh, 2007). Liu and Mattila (2017), indeed, posit that the match between the brand and the

96 Chapter 4: How and When consumer’s self-concept also affects consumers’ behavioral intentions. Consumers are more likely to respond positively toward a brand they feel connected to and that addresses their psychological needs (Aaker & Lee, 2001; Escalas & Bettman, 2003; Rajabi, Dens, & De Pelsmacker, 2015). Therefore, we expect:

H10: Consumers who have a stronger self-brand connection will have a more positive a) brand attitude, b) click intention and c) word-of-mouth intention.

The Mediating Role of Reactance to the Advertisement

Personalized advertising may also become too personal and may trigger psychological reactance from consumers, which is a state in which a person feels his/her freedom is threatened (Miron & Brehm, 2006; White et al., 2008). When confronted with a personalized advertisement, the consumer will try to restore his freedom either passively, by ignoring the advertisement, or actively, by behaving oppositely to what is intended by the advertiser (Miron & Brehm, 2006; White et al., 2008). For example, White et al. (2008) show that reactance to the advertisement decreases click-intention. Previous 4 research has found that consumers try to resist advertisements that are too personalized and will respond negatively to it (De Keyzer, Kruikemeier, et al., 2019; van Doorn & Hoekstra, 2013).

Perceived personal relevance. Perceived personal relevance might help to explain the link between personalized advertising and reactance to the advertisement. Personalized advertising can increase perceived personal relevance and as such, can decrease reactance to the advertisement due to its perceived utility. White et al. (2008) found that when an advertisement is perceived as useful for a consumer, consumers do welcome personalized messages and as such their advertisement avoidance decreases. We expect that personalized advertising will lead to a decrease in reactance to the advertisement via the indirect effect of perceived personal relevance.

H11: Consumers who perceive an advertisement on a social networking site as more personalized will express less reactance toward the advertisement via the mediating effect of perceived personal relevance.

Perceived entertainment. Perceived entertainment might also be helpful in explaining the link between personalized advertising and reactance to the advertisement. Personalized advertising can

97 Chapter 4: How and When increase perceived entertainment and, as such, can decrease reactance to the advertisement by its entertaining value. For example, Edwards, Li, and Lee (2013) demonstrate that consumers who perceived an advertisement as more entertaining are less likely to avoid the advertisement due to the fact that it brings them value. Therefore, we expect:

H12: Consumers who perceive an advertisement on a social networking site as more personalized will express less reactance toward the advertisement via the mediating effect of perceived entertainment.

Perceived intrusiveness. In study 1, we found that perceived personalization decreases intrusiveness. The decrease in intrusiveness will also decrease reactance to the advertisement and consumers will less likely behave opposite to what was intended. We, therefore, expect that personalized advertising will lead to less reactance to the advertisement via a decrease in perceived intrusiveness.

H13: Consumers who perceive an advertisement on a social networking site as more personalized will express less reactance toward the advertisement via the mediating effect of perceived intrusiveness.

Reactance theory (Miron & Brehm, 2006) suggests that consumers do not want to be manipulated and want to maintain their freedom. As such, they try to resist persuasion attempts, such as advertising, when confronted with them (Boerman & Kruikemeier, 2016). They may ignore the personalized ads, they may reject the ads or they might even try to find ways to block advertisement (Brinson, Eastin, &

Cicchirillo, 2018). Therefore, we expect:

H14: Consumers who have a stronger reactance toward the advertisement will have a more negative a) brand attitude, b) word-of-mouth intention and c) click intention.

Pretest

We conducted a pretest (N = 17, Mage = 23, SDage = 7.951, 76.5% female) to select the personalization elements for Study 2. To achieve a greater variance in perceived personalization, we sought to combine different personalization elements in a single advertisement in a stepped approach. Each respondent rated sixteen advertisements for a watch that used different personalization elements (life events, gender, age, interests, friends’ referrals, and retargeting) in different combinations, including one non- personalized control condition (see Table 4.3.). Chapter 2 already shows that the first five elements indeed contribute to perceptions of personalization. We added retargeting (personalization based on

98 Chapter 4: How and When prior online (search) behavior) to test another cue. This choice benefits the generalizability of the findings. Table 4.4. shows the manipulations of each separate element.

Table 4.3. Pretest: average perceived personalization per advertisement

Perceived Personalization scores Advertisement Mean SD Non-personalized 2.000 .748 Friends’ referrals 2.059 .833 Life events 2.224 .860 Retargeting 2.647 1.276 Gender and age 2.741 1.072 Interests 3.024 .935 Combination of life events and friends’ referrals 2.412 .862 Combination of gender, age, and life events 2.859 .929 Combination of gender, age, and friends’ referrals 2.859 1.214 Combination of gender, age, and interests 3.306 .791 Combination of interests and life events 3.435 .831 Combination of interests and friends’ referrals 3.219 .865 Combination of gender, age, interests, and friends’ referrals 3.141 .720 Combination of gender, age, interests, and life events 3.506 .803 Combination of gender, age, life events, and friends’ referrals 3.0824 1.022 Combination of gender, age, interests, life events, and friends’ 3.682 .7418 referrals

Table 4.4. Overview of manipulations per personalization element (translated from Dutch) 4 Condition Manipulations Slogan: What time is it? Time to discover why everyone is a fan of the top Non-personalized watches from Radar. Slogan: What time is it? Time to discover why Personalization based on [male/female][teenagers/people in their twenties/people in their thirties/…] gender and age are fans of top watches from Radar. Slogan: What time is it? Time to discover why others who like you, [enjoy Personalization based on playing sports/reading/gaming/shopping/ going out/ love nature/love interests culture] are a fan of the top watches from Radar. Above the advertisement: Others who recently [started a new training / Personalization based on life studied abroad / graduated / started a new job / started a new relationship / events ended their relationship / got engaged / got married / moved / got a new roommate / started a new hobby / traveled.] Personalization based on friend Above the advertisement: 3 of your friends like Radar. referrals

Respondents were recruited via Facebook. They were instructed to attentively look at the advertisements and rate the perceived personalization (Kalyanaraman & Sundar, 2006; Srinivasan et al., 2002) for each ad on a 5 point scale (1 = totally disagree, 7 = totally agree). We selected the non- personalized advertisement (1) as a control. We then sought to gradually increase the perceived personalization by adding one personalization element at a time. Because a personalization on interests and life events only did not significantly increase the perceived personalization over the control condition, we started with a personalization based on gender and age (2). This second condition was

99 Chapter 4: How and When significantly different from the control condition (t(16)= 2.963), p= .009). We refrained from using retargeting because retargeting is not usually combined with other elements. The addition of interests

(3) further increases the perceived personalization significantly (t(16)=2.934), p= .029). In the next steps, we added, consecutively, life events (4) and friend referrals (5). While these did not significantly increase perceived personalization over the combination of gender, age, and interests, the inclusion of many different personalization bases could lead to more irritation.

Main Study

Design and Stimuli

We set up a five-level between-subjects experiment. Table 4.5. provides an overview of the manipulations per condition. The advertisement was for a fictitious watch brand (Radar) and showed a unisex watch.

Table 4.5. Overview of manipulations per condition

Condition Manipulations Slogan: What time is it? Time to discover why everyone is a fan of Radar Non-personalized watches! Slogan: What time is it? Time to discover how other [men/women] [in their Combination of gender and age twenties/thirties] are also a fan of Radar watches! Slogan: What time is it? Time to discover how other [men/women] [in their Combination of gender, age twenties / thirties] who, like you, also like [socializing/ traveling/ sports/ and interests golf/ art and culture/ camping/ music/ hiking/ skiing/ skiing/ reading/ gaming/ nature/ shopping/ going out] are also a fan of Radar watches! Above the advertisement: Others who recently [got a new job/ started at a new school/ studied abroad/ started a new relationship/ got engaged/ got married/ celebrated an anniversary/ got a new child/ moved/ bought a Combination of gender, age, home/ travelled] also like Radar. interests and life events Slogan: What time is it? Time to discover how other [men/women] [in their twenties / thirties] who, like you, also like [socializing/ traveling/ sports/ golf/ art and culture/ camping/ music/ hiking/ skiing/ skiing/ reading/ gaming/ nature/ shopping/ going out] are also a fan of Radar watches! Above the advertisement: 1) Others who recently [got a new job/ started at a new school/ studied abroad/ started a new relationship/ got engaged/ got married/ celebrated an anniversary/ got a new child/ moved/ bought a home/ travelled] also like Combination of gender, age, Radar. interests, page-likes, and life 2) 3 of your friends also like Radar. events Slogan: What time is it? Time to discover how other [men/women] [in their twenties / thirties] who, like you, also like [socializing/ traveling/ sports/ golf/ art and culture/ camping/ music/ hiking/ skiing/ skiing/ reading/ gaming/ nature/ shopping/ going out] are also a fan of Radar watches!

In the non-personalized advertisement, the tagline (above the picture) read “What time is it? Time to discover why everyone is a fan of the top watches from Radar.”, which was the same to the one in the pretest. In the other conditions, we inserted respondents’ gender and age (and interests, depending on

100 Chapter 4: How and When the condition) into the tagline. In those conditions that also included life events and/or friend referrals, the information was inserted just above the advertisement, like in a real Facebook ad (see Appendix 1.D. for an example). The advertisements were embedded in a fictitious Facebook news feed, similar to the one used in Study 1, which contained one filler item (a profile picture update).

Procedure

Participants (N = 264, Mage = 28.011, SDage = 4.094, 53% male) were recruited using the crowdsourcing website Prolific. Based on a pre-screening, only panel members that used Facebook at least once a month were selected. As a verification, respondents indicated which social media they used at least once a month. Next, socio-demographic information was collected (gender, birth-year, degree). Then, respondents were asked to indicate the topics they were most interested in out of a list of 14 possibilities

(see Table 4.6. for the complete list). On the next page, we asked them about their Facebook use

(amount of days and minutes per day) and attitude toward Facebook. Subsequently, respondents were randomly allocated to one of the five conditions. We asked them to imagine that the Facebook news 4 feed they were about to see was their own and to look at the page attentively. After completing brand attitude, positive word-of-mouth intention, click intention, self-brand connection, perceived personal relevance, perceived intrusiveness, perceived entertainment, reactance to the advertisement, perceived personalization and perceived privacy protection by the social networking site measures, respondents were thanked for their participation. All items were measured using seven-point Likert scales or bipolar semantic differentials (see Table 4.6.).

Table 4.6. Measures (Study 2) Measure Items Cronbach’s Mean Source alpha (SD) Age Calculated based on: 28.011 What year were you born in? (4.094) Gender Are you  Male 53%  Female 47% Interests Please select the topics from the list below that you are most interested in.  Socializing  Traveling  Sports  Golf  Art and Culture  Camping  Music  Hiking  Skiing

101 Chapter 4: How and When

 Sports  Reading  Gaming  Nature  Shopping  Going out Perceived  This ad is tailored to my situation. .993 3.043 Srinivasan, personalization  I believe this ad is customized to (1.600) Anderson and my needs. Ponnavolu  This ad was targeted at me as a (2002) and unique individual. Kalyanaraman  I believe that this ad is customized and Sundar to my characteristics. (2006)  This ad was personalized according to my profile. Perceived personal  Not important – Important .923 3.052 De Keyzer et relevance  Not relevant – Relevant (1.569) al. (2015)  Meaningless - Meaningful Perceived  Distracting .937 4.442 Li, Edwards intrusiveness  Forced (1.73) and Lee  Interfering (2002) and  Intrusive Kim and Han  Invasive (2014)  Obtrusive  Irritating  Annoying Attitude toward the  Facebook is part of my everyday .819 4.182 De Keyzer et social networking site activity. (1.659) al. (2015)  I feel out of touch when I haven’t logged onto Facebook for a while.  I would feel sorry if Facebook shut down Perceived privacy  Facebook has adequate security .911 3.196 Wolfinbarger protection by the features. (1.360) and Gilly social networking site  I feel safe in my transactions with (2003) Facebook.  I feel like my privacy is protected at Facebook.  I trust Facebook will not misuse my personal information. Perceived The ad was .916 3.205 Kim and Han entertainment  Enjoyable (1.401) (2014) and  Pleasant Burns and  Entertaining Lutz (2006) Self-brand connection  [Brand] reflects who I am. .967 2.321 Escalas and  I can identify with [brand] (1.326) Stern (2003)  I feel a personal connection with [brand].  I could use [brand] to communicate who I am to other people.  I think [brand] would help me become the type of person I want to be.  I consider [brand] to be “me” (it reflects who I consider myself to be or the way I want to present myself to others)  ‘[Brand] suits me well.’ Reactance to the  I want to resist the advertisement. .876 5.924 Bleier and advertisement  I want to dismiss the content of this (1.414) Eisenbeiss advertisement. (2015)  I want to avoid this kind of advertisement.

102 Chapter 4: How and When

Brand attitude  Unappealing – Appealing .950 3.954 Spears and  Bad – Good (1.414) Singh (2004)  Unpleasant – Pleasant  Unfavorable - Favorable Positive word-of-  I am likely to say negative/positive .871 3.765 De Keyzer, mouth intention things about [brand] to other (1.220) Dens and De people. Pelsmacker  I am likely to discourage/encourage (2017) friends and relatives to buy [brand]. Click intention  It is not likely/likely that I will click 2.120 Bleier and this ad. (1.710) Eisenbeiss (2015) Facebook days  In the past week, on average, 56.4% De Keyzer, approximately how many days have used Dens and De you used Facebook? Facebook Pelsmacker every day (2017) Time spent on  In the past week, on average, how 50.8% De Keyzer, Facebook many minutes per day have you used Dens and De spent on Facebook? Facebook Pelsmacker less than (2017) 30 minutes per day Results

To replicate the findings of Study 1, we used Hayes’ (2018) PROCESS macro (model 4) with 5,000 bootstrap samples (see Table 4.7.). in line with Study 1, perceived personalization was entered as the 4 independent variable. Perceived personal relevance and perceived intrusiveness acted as mediating variables. We ran three distinct analyses with brand attitude (Ab), the click intention (CI), and positive word-of-mouth intention (WOM) as the dependent variables. As in Study 1, we report the completely standardized indirect effects.

Table 4.7. Unstandardized regression weights with perceived personal relevance and perceived intrusiveness as mediators (Study 2)

Perceived personal Perceived Ab WOM CI relevance intrusiveness B Sig b Sig b Sig b Sig b Sig Constant 2.695 <.001 3.800 <.001 4.517 <.001 3.775 <.001 1.792 .006 Perceived .457 <.001 -.149 .043 .108 .076 .019 .710 .253 <.001 personalization Perceived .290 <.001 .228 <.001 .478 <.001 personal relevance Perceived -.339 <.001 -.099 .048 -.229 <.001 intrusiveness Gender -.117 .503 -.094 .658 -.157 .248 .007 .961 -.117 .450 Age -.031 .145 .044 .063 -.001 .934 -.013 .449 -.026 .178 R² .220 .030 .433 .146 .491 The indirect effects of perceived personalization via perceived personal relevance on brand attitude (b

= .150, CI = [.077; .238], WOM (b = .136, CI = [.066; .216]) and click intention (b = .204, CI = [.131,

103 Chapter 4: How and When

285]) are positive and significant, reconfirming H12. The indirect effects of perceived personalization via perceived intrusiveness on brand attitude (b = .057, CI [.002; .118], and on click intention (b= .032,

CI = [.002; .065] are also positive and significant, reconfirming the unexpected findings of Study 1. The indirect effect of perceived personalization via perceived intrusiveness on WOM (b = .019, CI = [-.001;

.057]) was not significant. The indirect effects via perceived personal relevance on WOM intention

(brelevance – bintrusiveness = .117, [CI: .035; .205]) and on click intention (brelevance – bintrusiveness = .172, [CI:

.094; .263]) are significantly stronger than those through perceived intrusiveness. For brand attitude, the indirect effects through perceived personal and perceived intrusiveness were not significantly different (brelevance – bintrusiveness = .093, [CI: -.010; .198]).

Investigating Boundaries

We used Hayes’ (2018) PROCESS macro (model 59) with 5,000 bootstrap samples to test H3 – H6.

Perceived personalization served as the independent variable. Perceived personal relevance and perceived intrusiveness were the mediating variables. For H3 – H4, the attitude toward the social networking site was the moderator. In a separate analysis, we used the perceived privacy protection by the social networking site as the moderator to test H5 – H6. The dependent variables were brand attitude, positive word-of-mouth intention and click intention (in separate analyses). All variables were mean centered for the calculation of the interaction effects.

2 Note that the results from the analysis with actual personalization (corresponding to our manipulations, entered as a dummy variable; the control condition served as the reference category) as the independent variable (see Appendix 3.C.). The results show that the actual personalized conditions do not significantly affect neither perceived relevance (Bgender + age = -.171, p = .549; Bgender + age + interests = .344, p = .263; Bgender + age + interests + life-events = .353, p = .288; Bgender + age + interests + life events + friend referrals = .426, p = .171), nor perceived intrusiveness (Bgender + age = - .054, p = .873; Bgender + age + interests = -.041, p = .911; Bgender + age + interests + life-events = -.392, p = .275; Bgender + age + interests + life events + friend referrals = -.139, p = .696).

104

Table 4.8. Unstandardized regression weights with attitude toward the social networking site as moderator (Study 2)

Perceived personal relevance Perceived intrusiveness AB WOM CI b b b b Constant 1.308* -1.255 4.630*** 4.356*** 3.316*** Perceived personalization .461*** -.150* .127* .039 .267*** Perceived personal relevance .260*** .199** .450*** Perceived intrusiveness -.336*** -.095° -.221*** Attitude toward SNS .112* -.062 .137** .094° .105* Perceived personalization * Attitude toward SNS .065* -.025 .040 .054* .051° Perceived personal relevance * Attitude toward SNS -.010 -.007 .028 Perceived intrusiveness * Attitude toward SNS .013 .016 .028 Gender -.235 -.031 -.306* -.102 -.231 Age -.035° .047° -.008 -.016 -.031 R² .250 .036 .464 .179 .515 Notes: ***p≤ .001, ** p≤ .010, * ≤.050, ° ≤ .100

Table 4.9. Unstandardized regression weights with perceived privacy protection by the social network site as moderator (Study 2)

Perceived personal relevance Perceived intrusiveness AB WOM CI b b b b b Constant 1.046° -1.254° 4.317*** 4.085*** 2.999*** Perceived personalization .404*** -.063 .077 -.004 .209** Perceived personal relevance .272*** .198*** .438*** Perceived intrusiveness -.316*** -.096° -.205*** Perceived privacy protection by SNS .184** -.360*** .147* .078 .142* Perceived personalization * Perceived privacy protection by SNS .060° .034 .015 .083* -.020 Perceived personal relevance * Perceived privacy protection by SNS .017 .031 .147*** Perceived intrusiveness * Perceived privacy protection by SNS .040 .060° .035 and When 4: How Chapter Gender -.103 -.066 -.159 .033 -.075 Age -.033 .047* -.004 -.015 -.029 R² .255 .104 .456 .206 .544 Notes: ***p≤ .001, ** p≤ .010, * ≤.050, ° ≤ .100

105

Chapter 4: How and When

Table 4.8. shows the results from the analysis with attitude toward the social networking site as a moderator. First, we can again confirm the mediating roles of perceived personal relevance and perceived intrusiveness. H3 hypothesized that the effect of perceived personalization via perceived personal relevance on the dependent variables would be stronger with a more positive attitude toward the social networking site.

The conditional indirect effects of perceived personalization via perceived relevance on AB, WOM, CI are all significant. Figure 4.4. illustrates this effect for click intention. Moreover, the pairwise contrasts between the conditional indirect effects show that, at a 90% confidence level, the indirect effect on click intention via perceived relevance is significantly smaller for consumers with a more negative attitude towards the social networking site than for consumers with, respectively, a moderate (bnegative = .142, bmoderate = .208, CI = [.006, .123]) or more positive attitude towards the social networking site are not significantly different from each other (CI = [-.001, .162]. For the indirect effect of perceived personalization via perceived relevance on AB and WOM, no significant differences between the indirect conditional effects were found. Therefore, H3b, but not H3a or H3b, can be confirmed.

The conditional indirect effects of perceived personalization via perceived intrusiveness on AB, WOM and CI are only significant for consumers with a moderately positive attitude toward the social networking site. Moreover, the conditional indirect effects are not significantly different from each other.

Therefore, H4 cannot be confirmed.

H5 hypothesized that the effect of perceived personalization via personal relevance on the dependent variables would be stronger when users’ perceived privacy protection by the social networking site is higher (Table 4.9.). The conditional indirect effects of perceived personalization via personal relevance on brand attitude are all significant. The conditional indirect effects via personal relevance on CI are significantly different for the different levels of perceived privacy protection by the social networking site (Figure 4.5. illustrates this effect for click intention). More specifically, the indirect effect of personalization via personal relevance on click intention becomes stronger when users’ perceived privacy protection by the social networking site is higher. The difference between the indirect effect of perceived personalization via perceived relevance on click intention for consumers with a lower level of perceived privacy protection by the social networking site (b = .077) and a moderate level of perceived privacy

106 Chapter 4: How and When protection by the social networking site (b = .177) is significantly different (CI = [.061, .142]. This is also true for the difference between the indirect effect for consumers with a lower level of perceived privacy protection and a higher level of perceived privacy protection (b = 310, CI = [.129, .340] and for the difference between moderate and higher levels of perceived privacy protection (CI = [.066,

.201]). Therefore, H5c is confirmed. The conditional indirect effects for brand attitude and WOM are not significantly different and, thus, H5a and H5c are not confirmed.

Figure 4.4. Conditional indirect effects of perceived personalization via perceived relevance on click intention at different levels of perceived privacy protection by the SNS

Notes: LLCI = lower limit confidence interval; ULCI = upper limit confidence interval

H6 hypothesized that the indirect effect via perceived intrusiveness would be stronger when users’ perceived privacy protection by the social networking site is higher. The conditional indirect effects were not significant at any level of perceived privacy protection and the pairwise contrasts are also not significant. Therefore, H6 is not supported.

Extending the Processing Mechanism

To analyze the hypothesized extended processing mechanisms of perceived personalization (H7- H14), we tested three structural equation models (one for each dependent) in SmartPLS3. The model fit was

107 Chapter 4: How and When acceptable for both Ab (SRMR = .054, χ² = 519.865, NFI = .948), WOM intention (SRMR = .057, χ² =

506.731, NFI = .943) and CI intention (SRMR = .055, χ² = -200.699, NFI = 1.023). The factor loadings for all indicators are sufficiently large and significant, indicating convergent validity (Table 4.10.).

Table 4.10. Measurement model (Study 2)

Construct Item Factor loadings Composite AVE AB WOM Sig. reliability model model Pers_1 1.018 1.020 <.001 Pers_2 .503 .507 <.001 Perceived Pers_3 .822 .819 <.001 .934 .750 personalization Pers_4 .813 .814 <.001 Pers_5 1.060 1.057 <.001 Rel_1 .921 .914 <.001 Perceived personal Rel_2 .893 .886 <.001 .924 .802 relevance Rel_3 .872 .886 <.001 Int_1 .511 .499 <.001 Int_2 .795 .773 <.001 Int_3 .762 .772 <.001 Perceived Int_4 .791 .799 <.001 .935 .652 intrusiveness Int_5 .737 .728 <.001 Int_6 .712 .730 <.001 Int_7 1.022 1.020 <.001 Int_8 1.015 1.022 <.001 Ent_1 .928 .903 <.001 Perceived Ent_2 .912 .873 <.001 .918 .790 entertainment Ent_3 .824 .885 <.001 Con_1 .902 .893 <.001 Con_2 .955 .945 <.001 Con_3 .921 .952 <.001 Self-brand Con_4 .891 .888 <.001 .968 .810 connection Con_5 .862 .862 <.001 Con_6 .873 .904 <.001 Con_7 .894 .854 <.001 Reac_1 .657 .654 <.001 Reactance to the Reac_2 .878 .857 <.001 .879 .713 Advertisement Reac_3 .968 .991 <.001 AB_1 .931 <.001 AB_2 .903 <.001 Brand attitude .952 .831 AB_3 .872 <.001 AB_4 .939 <.001 WOM WOM_1 .869 <.001 .873 .775 WOM_2 .891 <.001

108

Table 4.11. Square root of average variance extracted and correlation per construct

Perceived Perceived Perceived personal Perceived Self-brand Reactance to the Dependent personalization entertainment relevance intrusiveness connection advertisement AB WOM AB WOM AB WOM AB WOM AB WOM AB WOM AB WOM Perceived .866 .866 personalization Perceived .532 .533 .889 .887 entertainment Perceived personal .511 .511 .668 .671 .895 .896 relevance

Perceived intrusiveness -.169 -.170 -.583 -.583 -.375 -.375 .808 .808

Self-brand connection .568 .568 .687 .689 .633 .633 -.347 -.347 .900 .901 Reactance to the -.357 -.357 -.655 -.656 -.539 -.538 .673 .672 -.559 -.558 .844 .845 advertisement Dependent .366 .222 .722 .413 .555 .396 -.579 -.284 .640 .491 -.616 -.385 .912 .880 Notes: ***p≤ .001, ** p≤ .010, * ≤.050, ° ≤ .100

Table 4.12. Unstandardized regression coefficients for brand attitude, positive word-of-mouth intention and click intention (Study 2: Extending the Mechanism)

Perceived Perceived Perceived Self-brand Reactance to the AB WOM CI entertainment relevance intrusiveness connection advertisement b b b b b b b b Perceived personalization .532*** .511*** -.169* .223*** -.063 -.078 -.125 .088 Perceived entertainment .417*** -.238* .352*** .054 .166°

Perceived relevance .251** -.176* .052 .115 .280*** and When 4: How Chapter Perceived intrusiveness .027 .458*** -.214*** -.034 -.092° Self-brand connection .292*** .389*** .227** Reactance to the -.079 -.093 -.079 advertisement Notes: ***p≤ .001, ** p≤ .010, * ≤.050, ° ≤ .100

109

Chapter 4: Let’s get personal

Moreover, the average variance extracted (AVE) for each factor was above the .50 threshold.

Furthermore, reliability estimates range between .873 and .968, which are well above the recommended

.70 (Hair, Black, Bain, & Anderson, 2014). In Table 4.11., the diagonals show the square root of the

AVE per construct and the off-diagonals show the correlation between each pair of constructs. No correlation was found to be higher than the square root of the AVE, confirming discriminant validity

(Fornell & Larcker, 1981).

Table 4.12. shows the unstandardized regression coefficients for brand attitude, positive word-of-mouth intention and click intention. We find a significant, positive indirect effect of perceived personalization via perceived personal relevance on self-brand connection (bAB = .132, p = .001; bWOM = .127, p = .001; bCI = 129, p = .003), which confirms H7. Next, we also find a significant, positive indirect effect of perceived personalization via perceived entertainment on self-brand connection (bAB = .218, p < .001; bCI = 218, p < .001; bWOM = .223, p <.001). H8 is supported. We also hypothesize an indirect effect of perceived personalization on self-brand connection via perceived intrusiveness. However, our findings do not confirm H9: (bAB = -.005, p = .714; bCI = -.004, p = .687; bWOM = -.004, p =.705). Self-brand connection was found to positively impact brand attitude (b = .292, p <.001), click intention (b = .227, p = .016) and WOM (b = .389, p < .001). H10 is confirmed.

We find a significant, negative indirect effect of perceived personalization on reactance to the advertisement via perceived personal relevance (bAB = -.091, p = .028; bCI = -.090, p = .020; bWOM = -

.089, p = .015), confirming H11. We also find a significant, negative indirect effect of perceived personalization on the reactance to the advertisement via perceived entertainment (bAB = -.125, p =

.031; bCI = -.126, p = .017; bWOM = -.130, p = .019), which confirms H12. Perceived intrusiveness also acts as a mediating variable between the relationship of perceived personalization and reactance tot the advertisement (bAB = -.078, p = .024; bCI = -.076, p = .019; bWOM = -.077, p = .016). H13 is supported.

Finally, we hypothesized that reactance to the advertisement would negatively affect consumer responses. However, the effects on Ab (b = -.079, p = .361), click intention (b = -.079, p = .364) and

WOM (b = -.093, p = .448) are not significant. H14 cannot be confirmed.

110 Chapter 4: Let’s get personal

General Discussion and Conclusion

Based on two experiments, the present study examines the processing mechanism of personalized advertising. Table 4.13. provides an overview of the hypotheses and their conclusions. Our findings confirm previous research on the positive explanatory mechanism of perceived personal relevance (De

Keyzer et al., 2015; Kim & Huh, 2017). Contrary to our expectations, we find that perceived personalization decreases perceived intrusiveness, which results in a more positive brand attitude, click intention and WOM intention. Based on White et al. (2008) and Young and Kim (2019) we initially expected a negative indirect effect. On the other hand, Kim and Han (2014) and Ketelaar et al. (2018) also reported this positive effect. They argue that personalized advertisements entail a greater value for consumers and, as such, are less likely to be perceived as unwanted. Pfiffelmann et al. (2019) did not find a direct effect of personalization on perceived intrusiveness of the advertisement, but the effect was mediated by visual attention. This increase in visual attention helped to reduce the attitudinal persuasion knowledge. Learning effects could explain why more recent research does not replicate previously documented negative indirect effects of personalization through intrusiveness. The 4 Persuasion Knowledge Model (PKM) (Friestad & Wright, 1994) posits that people learn to cope with persuasive attempts: they learn their persuasive intent and they learn to recognize these persuasive attempts. As such, social networking site user might have learned that there are advertisements in their social networking sites and how to recognize them. Therefore, they might not feel as intruded when they come across an advertisement as users would have when social networking sites were less familiar.

We also set out to examine the processing mechanism by exploring its boundary conditions. To our knowledge, no previous study has examined the moderated mediation which was proposed here. First, we only found a moderating effect of the attitude toward the social networking site on the indirect effects of perceived personalization on click intention, but not on brand attitude or positive word-of- mouth intention, via perceived personal relevance. We did not find moderating effect of attitude toward the social networking site on the indirect effects of perceived personalization on consumer responses via perceived intrusiveness.

111 Chapter 4: Let’s get personal

Table 4.13. Overview of the hypotheses

Hypothesis Conclusion H1: Consumers who perceive an advertisement on a social networking site as Supported. more personalized will have a more positive brand attitude, via the mediating effect of perceived personal relevance of the advertisement. H2: Consumers who perceive an advertisement on a social networking site as Not supported. more personalized will have a more negative brand attitude, via the mediating effect of perceived intrusiveness of the advertisement. RQ1: What is the relative strength of the mediating effects of perceived personal The indirect effect relevance and perceived intrusiveness on the relation between ad personalization through personal and brand attitude? relevance is stronger for AB, CI and WOM H3: The positive effect of perceived personalization via perceived personal Not supported. relevance on a) positive word-of-mouth and c) click intention is stronger when the attitude toward the social networking site is more positive. H4: The positive effect of perceived personalization via perceived intrusiveness on Not supported. a) positive word-of-mouth and c) click intention is stronger when the attitude toward the social networking site is more positive. H5: The positive effect of perceived personalization via perceived personal H5a and H5c: not relevance on a) positive word-of-mouth and c) click intention is stronger when supported. users’ perceived privacy protection by the social networking site is higher. Hb: supported. H6: The positive effect perceived personalization via perceived intrusiveness on a) Not supported. positive word-of-mouth and c) click intention is stronger when users’ perceived privacy protection by the social networking site is higher. H7: Consumers who perceive an advertisement on a social networking site as Supported. more personalized will have a stronger self-brand connection via the mediating role of perceived personal relevance. H8: Consumers who perceive an advertisement on a social networking site as Supported. more personalized will have a stronger self-brand connection via the mediating role of perceived entertainment. H9: Consumers who perceive an advertisement on a social networking site as Not supported. more personalized will have a stronger self-brand connection via the mediating role of perceived intrusiveness. H10: Consumers who have a stronger self-brand connection will have a more Supported. positive a) brand attitude, b) word-of-mouth intention and c) click intention. H11: Consumers who perceive an advertisement on a social networking site as Supported. more personalized will express less reactance toward the advertisement via the mediating effect of perceived personal relevance. H12: Consumers who perceive an advertisement on a social networking site as Supported. more personalized will express less reactance toward the advertisement via the mediating effect of perceived entertainment. H13: Consumers who perceive an advertisement on a social networking site as Supported. more personalized will express less reactance toward the advertisement via the mediating effect of perceived intrusiveness. H14: Consumers who have a stronger reactance toward the advertisement will Not supported. have a more negative a) brand attitude, b) word-of-mouth intention and c) click intention.

Perceived privacy protection by the social networking site was examined as a second potential moderator. Previous research has indicated that by creating trust, which can be done by being clear about privacy protection measures or providing user control, a spillover-effect from the website to the embedded, personalized, advertisements can take place (Aguirre et al., 2015). Moreover, Mpinganjira and Maduku (2019) show that perceived privacy control has a negative influence on privacy concerns and, indirectly, on ad avoidance. When users are confident their privacy is protected, they may

112 Chapter 4: Let’s get personal appreciate personalized advertising more. As a result, they might value the personal relevance of the personalized advertisement more highly which results in positive brand responses. Our findings indicate that this is only true for click intention and for the indirect effect through perceived relevance. More specifically, the effect of perceived personalization on click intention through perceived personal relevance grows stronger with increasing levels of perceived privacy protection.

The final contribution of the current study lies in the extension of the processing mechanism of personalized advertising. We examined the antecedents of self-brand connection via the indirect effects of perceived personalization through perceived personal relevance, perceived entertainment, and perceived intrusiveness. Our findings suggest that perceived personalization has a positive effect on self-brand connection via perceived personal relevance and perceived entertainment, but not via perceived intrusiveness. This relates to the self-referencing process triggered by perceived personalization, both via relevance and entertainment it is clear that consumers’ psychological needs are addressed. However, perceived intrusiveness does not hinder the self-referencing process. Our 4 findings also indicate that perceived personalization can decrease reactance to the advertisement via perceived entertainment, perceived relevance, and perceived intrusiveness. Edwards et al. (2013) also find that perceived entertainment can decrease the psychological reactance toward a personalized advertisement due to the value that it generates for consumers. Contrary to our expectations reactance to the advertisement did not significantly affect brand responses. This might be explained by the findings of Marotta, Zhang, and Acquisti (2015), who noted that while consumers do not like being tracked online, they often appreciate the benefits of online targeted advertising. As a result, they will not be inclined to try to react negatively to the advertisement.

In line with Chapter 3, we again found that the direct effects of perceived relevance on brand attitude

(Study 1: β = .430, <.001; Study 2: β = .521, <.001), click intention (Study 1: β = .464, <.001;

Study 2: β = .631, <.001) and WOM (Study 2: β = .355, <.001) and of perceived intrusiveness on brand attitude (Study 1: β = -.355, <.001; Study 2: β = -.540, <.001) and WOM (Study 2: β = -.248,

<.001) are larger than the direct effects of perceived personalization on brand attitude (Study 1: β =

.229, <.001; Study 2: β = .323, <.001), click intention (Study 1: β = .339, <.001; Study 2: β = .466,

<.001) and WOM (Study 2: β = .178, <.001). This was not the case for the effect of perceived

113 Chapter 4: Let’s get personal intrusiveness on click intention (Study 1: β = -.290, <.001; Study 2: β = -.419, <.001). Therefore, both perceived relevance and perceived intrusiveness do not only have to do with the message being personalized based on personal characteristics. Other factors that might affect perceived relevance and intrusiveness could be the timing with which the message is delivered, the frequency and the channels used are also meaningful to consumers (McMurtry, 2017).

Managerial implications

Our findings also have implications for advertisers and marketers. In general, personalized advertising on social networking sites leads to more positive consumer responses. This is because personalized ads are perceived as more relevant, more entertaining and less intrusive. As a result, it becomes easier for a consumer to connect the brand to his own self-image, which results in a more positive brand attitude, word-of-mouth intention, and a higher click intention. As a result, both attitudes and behavioral intentions can be positively affected by personalized advertising. Moreover, the increase in self-brand connection might also foster more long-term effects: a self-brand connection results in stronger and more favorable brand associations in the consumers’ memory. As such consumers might become committed to brands with which they have a strong self-brand connection (Escalas, 2004; Escalas &

Bettman, 2003; Palazon et al., 2018). At the same time, it is important to not creep consumers out with levels of personalization that are too high, because this cost would outweigh the benefits of personalization causing adverse effects (De Keyzer, Kruikemeier, et al., 2019).

Our findings also indicate that brand managers should be aware of the effects of the context (i.e., the social networking site) in which they place their advertisements. In general, social networking sites that are more liked and that are more clear about the privacy protection measures can reinforce the positive effects of perceived personalization. More specifically, presenting a personalized advertisement on a social networking site that is better liked or perceived as protecting consumers’ privacy will result in higher positive word-of-mouth intention. Second, presenting a personalized advertisement on a social networking site which is at least moderately liked by the consumer or is perceived by the consumer as protecting his privacy, will also reinforce the positive indirect effect of personal relevance and thus will result in higher click intentions.

114 Chapter 4: Let’s get personal

Finally, our findings also have implications for social networking site owners. Based on our results, it seems a good practice to clearly communicate about privacy protection practices and to try to create a likable brand image because both practices strengthen the indirect effect of perceived personalization on click intention via perceived relevance. To make sure that the advertisements on the social networking site have the positive intended effects, social networking site owners could provide advertisers with guidelines to increase the likelihood that personalized advertisements create positive brand attitudes and higher click and word-of-mouth intentions.

Limitations and Future Research

Both experiments presented here investigate personalized advertising for more hedonic products (travel and a watch). Hedonic products are used because of their aesthetic or sensory experience, and for amusement, fantasy, or fun (Hirschman & Holbrook, 1982). They are typically evaluated on subjective characteristics (shape, taste, look) (Dhar & Wertenbroch, 2000). Future research should also examine utilitarian products (e.g., insurance) to test whether our findings generalize to more utilitarian products. 4 As utilitarian products are usually more cognitively evaluated (Dhar & Wertenbroch, 2000), the more evaluative mediators (relevance and intrusiveness) might gain more importance. Moreover, De Keyzer et al. (2015) already indicated that personalized advertisements for hedonic products might be more congruent with consumers’ motives for using SNSs than advertisements for utilitarian products. As a result, they might be more readily perceived as relevant and entertaining.

Second, we used Facebook as a context because it is currently the most popular social networking site in the world. However, Voorveld (2019) argues that consumer responses to brand communication in various types of social media platforms differ. In line with the context appreciation theory, the platform in which the personalized advertisement is embedded can also be used as a source of information in the processing mechanism of personalized advertising. In the context of television advertising, it positively influences likeability, informativeness and brand recognition (De Pelsmacker, Geuens, &

Anckaert, 2002). As such, future research is encouraged to test the current model in different social media platforms and to provide for an explicit comparison between platforms.

115 Chapter 4: Let’s get personal

Third, the create a better understanding of the effects of personal relevance and perceived intrusiveness future research could try to distinguish their effects by manipulating them separately. In doing that researchers could provide insights into the effects each of the variables has on consumers’ responses.

Next, perceived privacy protection by the social networking site might be a particularly important moderator for consumers with a high, pre-existing, privacy concern. In our sample the average self- reported privacy concern was rather high (M = 5.80, SD = 1.01). Therefore, we suggest future research to test the interaction between privacy concern and perceived privacy protection by the social networking site to gain further insights on the boundary conditions of perceived personalization.

Moreover, it is unclear what users of social networking sites perceive as being too close. For example,

Čaić, Mahr, Aguirre, de Ruyter, and Wetzels (2015) found that the use of location (retrieved from your smartphone) in personalized mobile advertising increases feelings of vulnerability impeding the potential positive effects of personalization. Moreover, in a working paper we also found that when consumers aren’t aware specific information (e.g. location shared over WI-FI or browsing behavior when not logged onto the social networking site) is shared, they find it creepy when that information is used in advertising on a social networking site (De Keyzer, Dens, & De Pelsmacker, 2019). We suggest future research to look into what is perceived as personal space by consumers when using social networking sites and where the tipping point between the relevance of using personal information and creepiness of using personal information lies. Finally, the present study exposed participants to static mock Facebook news feeds. The resulting lack of interactivity could affect the results. Moreover, the external validity could be improved by performing a field experiment in a real, interactive social networking environment.

116

Going too far? How Consumers Respond to Personalized Advertising from Different Sources1,2

1 Manuscript under review as De Keyzer, F., van Noort, G., & Kruikemeier, S. Going too far? How Consumers Respond to Personalized Advertising from Different Sources. 2 Earlier versions of this chapter were presented at the 2018 Etmaal van de Communicatiewetenschap in Ghent, Belgium and at 2019 International Communication Association in Washington, D.C., United States. 5 Chapter 5: Going too far?

Abstract

This paper studies the impact of level of personalization (low vs. moderate vs. high) in advertisements from four online sources (health vs. governmental vs. commercial vs. news) on source attitudes. Based on the privacy calculus theory, we tested a moderated mediation model with perceived creepiness and relevance as competing mediating variables and source type as the moderating variable. The results of an experiment (N = 619) indicate that perceived creepiness negatively explains personalization effects and that the tipping point lies at the low level: a moderate (vs. low) level of personalization increases perceived creepiness, but high personalization does not increase it further. Contrary to our expectations, perceived relevance does not act as a positive explanatory mechanism. Finally, our findings demonstrate that source type is important in explaining personalization effects: the privacy calculus for each of the personalization levels was different for different online sources. Practical and research implications are discussed.

118 Chapter 5: Going too far?

Introduction

The majority of online ads are now personalized towards individual Internet users (IHS Markit, 2015), due to new technologies and changes in the media landscape, such as widely applied algorithms,

Artificial Intelligence (AI), and an upsurge of social media use, enabling the tracking and usage of consumer’s online information (both personal and social). Especially, Social Networking Sites (SNSs) are an extremely important venue for personalized advertising practices, because of its global adoption and the amount of personal information disclosed within these networks (De Keyzer et al., 2015; Kelly, Kerr,

& Drennan, 2017). It is often argued that advertising online will become even more personalized in the future (Kumar & Gupta, 2016; Schultz, 2016). Currently, personalization elements used in advertising differs greatly; consumers can be addressed based on their name, browsing behavior, social ties, groups and preferences, and all possible combinations of such personal information (Arora et al., 2008; De Keyzer et al., 2015; Hawkins et al., 2008; Hoy & Milne, 2010). In order to understand the implications of the algorithms used for personalization, we need to understand how people feel about them (Bucher,

2017). Although most people do not know exactly how such algorithms work, they try to make sense of them (i.e., so-called algorithmic imaginary; Bucher, 2017). Therefore, this study aims to investigate 5 how social media users respond to personalized advertisements that are the result of algorithms making use of different types of personal data, resulting in different levels of ad personalization.

The industry and scholars often assert that personalized advertising is – to some extent – more effective than non-personalized advertising (e.g., it is more memorable, attracts more attention, and sparks behavioral change; Malheiros et al., 2012; Noar et al., 2007; Sohl & Moyer, 2007). However, others found that personalized advertising may have a negative impact: personalized ads are perceived as creepy (e.g. Malheiros et al., 2012) or invasive (van Doorn & Hoekstra, 2013; White et al., 2008). This contradiction is also highlighted by Ur et al. (2012, p. 6), who argue that “taken as a whole, participants found online behavioral advertising3 smart, useful, scary and creepy at the same time”. Thus, while previous work has uncovered these contradictory feelings, three important gaps can be identified.

3 Online behavioral advertising can be seen as a particular form of personalized advertising as it uses online behavior to generate a personal profile of a consumer, which later on can be used to personalize advertising.

119 Chapter 5: Going too far?

First, underlying processes, both beneficial and detrimental, explaining the effectiveness of personalized advertising remain uncertain (Boerman et al., 2017). Positive consumer responses are usually explained by an increase in (perceived) relevance (De Keyzer et al., 2015), whereas negative responses are usually ascribed to a heightened level of privacy concern (Chellappa & Sin, 2005). This study extends previous work by deploying the privacy calculus theory to examine these competing responses to personalized advertising. Studies have applied privacy calculus theory in the context of self-disclosure behavior (e.g.

Culnan & Armstrong, 1999; Dienlin & Metzger, 2016; Laufer & Wolfe, 1977); however, privacy calculus might not only affect disclosure of personal information online but also affect other consumer responses such as attitudes towards the advertiser (Demmers et al., 2018). This study will build upon this work and examine how people weigh the benefits and the costs of personalized advertising in the context of social networking sites towards the advertiser, and how this translates into impact.

Second, this study examines two boundary conditions of the privacy calculus in two ways. A first boundary condition deals with the use of different personalization elements. Previous research indicated that personalization techniques can result in “over-personalized” advertisements, referring to ads that become too personalized. In that case, the advertisement contains too much personal information to the extent that it becomes creepy (Gironda & Korgaonkar, 2018). Therefore, this study is also breaking new ground as it examines more closely what the ‘tipping point’ is of personalization. According to

Malheiros (2013, p. 146) “advertisers should aim for sweet-spot personalization of ads”, which refers to the level of personalization that maximizes the noticeability and users’ conform level with personalization. The current study sheds light on that sweet spot by examining the impact of three different combinations of personalization elements in one design.

A second boundary condition is context-dependency. Although previous research has indicated that consumer responses toward online behavioral advertising depend on the context (e.g. Smit et al., 2014), studies examining the effects of personalized content or advertising in different contexts are rare (see e.g., Smit et al., 2014, and Ur et al., 2014, and Bol et al. 2018). In line with previous suggestions made in the literature, we expect that in some cases (e.g., in news websites), personalization is more acceptable than in other cases (e.g., health websites), because consumers have learned that the use of personal information is appropriate in some contexts, but not in others (Acquisti et al., 2015). Therefore,

120 Chapter 5: Going too far? the current study examines whether the impact of personalized ads is context-dependent, by investigating the effects for four different online sources. In sum, using an experimental design, we examine competing underlying mechanisms and the boundary conditions of personalized advertising. We conclude with management implications and a future research agenda.

Personalized Advertising: Explaining Effects Using the Privacy Calculus Theory The privacy calculus theory can be applied to uncover why some studies have found beneficial effects of personalization, whereas others have found negative effects. The theory posits that consumers weigh the perceived benefits (e.g., perceived relevance) with the perceived costs (e.g., perceived privacy invasion; (Culnan & Armstrong, 1999; Laufer & Wolfe, 1977). In case the benefits trump the costs, consumers will respond positively to personalized advertising. For example, when people are exposed to SNS advertisements that are more personalized on the basis of personal data, these ads become more relevant and useful (De Keyzer et al., 2015). It also keeps these sites free for users to use (Kelly et al., 2017). Both of these arguments can be considered as benefits of personalized advertising.

Conversely, large amounts of personal information is collected via these sites which can be considered 5 a cost of personalized advertising (Gironda & Korgaonkar, 2018; Li, Sarathy, & Xu, 2011), and this might increase feelings of creepiness (Malheiros et al., 2012; Ur et al., 2012). People are concerned about who is ‘snooping’ around in their personal social media profiles and what is known about them. Moreover, previous research indicates that personalized advertising increases privacy concern due to the increased risk of data theft and misuse of personal information (Baek & Morimoto, 2012; Ham, 2016). In sum, although consumers seem to be concerned about their privacy when they encounter personalized advertising (Malheiros et al., 2012), they might also appreciate the more relevance and usefulness of personalized advertisements (Ur et al., 2012) at the same time. These competing beliefs may override each other and as a result influence the outcomes of the privacy calculus (Dinev & Hart, 2006): some people might prefer personalized advertising as it increases relevance and decreases advertising clutter but others might be too concerned about the invasion of their privacy.

121 Chapter 5: Going too far?

The Cost Side of Personalized Advertising: Perceived Creepiness

Creepiness is a term originated from psychology and refers to “anxiety aroused by the ambiguity of whether there is something to fear or not and/or by the ambiguity of the precise nature of the threat that might be present” (McAndrew & Koehnke, 2016, p. 10). As such, it is not the actual presence of danger that causes feelings of creepiness, but the uncertainty of danger being present or not (McAndrew

& Koehnke, 2016). This perceived creepiness is also more and more used in non-psychological work, such as communication and marketing studies. In these fields, slightly different definitions have been used. For example, according to Malheiros et al. (2012, p. 581), perceived creepiness is “a sense that someone has been ‘snooping’ into a part of your life that should remain private.” Barnard (2014, p. 1) describes it similarly: “the marketer is watching her (ed. the consumer) and her privacy has been violated”. Moreover, Zhang and Xu (2016, pp. 1678-1679) state that, “[i]nstead of causing actual harms, in many privacy concerning cases, what novel technological features do is to trigger a sense of expectation violation or loss of control. In privacy research, creepiness is one particular emotional reaction to novel technological features, which is a mixture of fear, anxiety, and strangeness”. It is clear that the underlying factor is the same: the uncertainty and unpredictability of behavior with which one is confronted.

Uncertainty is also prevalent when consumers are confronted with personalized advertising. Previous studies have found that consumers are uncertain of how personal data, used in personalized advertising, is collected (Malheiros et al., 2012; Ur et al., 2012), and how these data are used in algorithms which results in feelings of creepiness (Bucher, 2017). People might wonder: ‘Is someone – or something – snooping around in my personal social media profile?’, ‘Are “they” following what I do on the Internet?’ and thus, ‘What do “they” know about me?’ As a result, people express a need for consent to collect data about them (Malheiros et al., 2012). When they do not provide such a consent, people feel uncomfortable, uncertain, and therefore perceive the personalized advertisement as creepy, especially when the level of personalization is too high (e.g., using a personal picture; Malheiros et al., 2012) .

This perception of creepiness might result in what White et al. (2008) call personalization reactance.

When the level of personalization is too high, consumers might react to it in a similar way as in case of psychological reactance. In such a case, consumers try to restore their freedom of choice and behave

122 Chapter 5: Going too far? opposite to what is intended by the threat (Miron & Brehm, 2006). Therefore, they will try to resist the advertising message and will respond negatively to it (van Doorn & Hoekstra, 2013; White et al., 2008). In our study they might do this by evaluating the message source as less positive. We expect that the use of more personalization elements (i.e., using more detailed personal information) will make it more likely that an individual perceives personalized ads as creepier and, consequently, evaluates the source less positively. Therefore, we expect:

H1: (a) The use of more personalization elements results in higher levels of perceived creepiness of the advertisement, and (b) subsequently, into less positive source attitudes.

The Benefit Side of Personalized Advertising: Perceived Relevance

Previous studies suggesting a positive impact of personalization and ad responses propose personal relevance as an explanation. A number of prior studies have established the mediating role of perceived relevance (e.g. De Keyzer et al., 2015; Kalyanaraman & Sundar, 2006; Sundar & Marathe, 2010). The

Elaboration Likelihood Model (Petty & Cacioppo, 1986) is often used to explain these positive effects.

Essentially, consumers who are confronted with a persuasive message might try to relate the message 5 to themselves, which is called self-referencing (Hawkins et al., 2008). Under peripheral processing, this self-referencing is used as a heuristic cue or a decision aid. For example, when the message is about women’s apparel, most men would not relate to this message. Under central processing, on the other hand, self-referencing could motivate consumers to process the message more elaborately which ultimately might lead to more positive evaluations of the message (Bright & Daugherty, 2012). As a result, under both peripheral and central processing, this self-referencing can result in positive responses

(Hawkins et al., 2008; Romeo & Debevec, 1992). For example, in Chapter 3 we found that even a basic level of personalization (based on gender) increases perceived relevance and results in a more positive attitude toward the source of the message (i.e., the brand) under both central and peripheral processing. Moreover, Chapter 4 reproduced this effect with different levels of personalization.

Therefore, we expect:

H2: (a) The use of more personalization elements results in higher perceived relevance of the advertisement, and (b) subsequently, into more positive source attitudes.

123 Chapter 5: Going too far?

In the two previous sections, we established that there might be two competing underlying mechanisms which are important when measuring the effects of levels of personalization on source attitudes.

Previous work on the privacy calculus shows that these mechanism work in competition but are weighted differently depending on the case or context which is studied. For instance, Dienlin and Metzger (2016) found that with regard to self-disclosure in SNSs, the benefits outweighed privacy concern, whereas, with regard to self-withdrawal, the privacy concern outweighed both privacy self-efficacy and the benefits. Also, Wang, Duong, and Chen (2016) show that in the context of mobile applications, the perceived benefits outweigh the perceived risks in predicting the intention to disclose personal information. In general, previous research indicates that both benefits and costs affect outcomes

(Dienlin & Metzger, 2016). Because of conflicting findings on how the mechanisms compete, we propose the following research question for personalized advertising in social networks:

RQ1: Which mediating variable (i.e., perceived creepiness and perceived relevance of the advertisement) is stronger in explaining the effects of personalization in advertising on source attitudes?

Investigating Boundaries: The moderating effect of source type

The use of personal information can be seen as appropriate in one context, but unacceptable in another as posited by Acquisti et al. (2015). They state that people have learned coping strategies over time, based on situational criteria. Therefore, depending on the situation, people have different perceptions on how others – or in this case algorithms – handle their personal information. For example, in a study by Ur et al. (2012) participants were willing to allow data collection when reading the news, but not when searching an STD treatment for a friend. According to Nissenbaum (2010), to determine whether the use of information is acceptable or not, the situation in which data transaction takes place should first be identified. Next, the values at stake, the contextual norms and expectations can be determined.

When a data transaction respects these values, norms, and expectations within a certain context, it is considered appropriate. For personalized advertising in social networking sites, such context might differ in terms of how personal the personal information is perceived. As is shown in the study of Ur et al.

(2012): health-related information is seen as much more personal and therefore less appropriate for use in advertising than for example information on which news articles are read. This clearly indicates that the same personal information is perceived differently in various contexts and thus that

124 Chapter 5: Going too far? personalization effects may differ in each context. Therefore, we expect that the impact of personalization through creepiness and relevance is moderated by the online source.

There is only some anecdotal evidence for the context-dependency of personalization effects: for example, Bol et al. (2018) find that personalization affected outcomes in a news and commercial context, but not in a health context. In the following section, we give a brief overview of studies examining personalization effects in these four contexts (health website, governmental website, online newspaper, and a commercial website) separately.

First, in a health context, it seems that the use of tailoring and personalizing persuasive health messages has a long history (e.g. Hawkins et al., 2008; Noar et al., 2009). These studies have often identified positive effects of personalization via relevance. As mentioned before, the use of personal information increases the likelihood of self-referencing, and as a result, increases perceived relevance. That, in turn, increases processing under both the central and the peripheral route, leading to more positive consumer responses. However, there is also evidence that personalized health-related messages are perceived as embarrassing or too personal (Barnard, 2014), resulting in higher levels of reactance. The main rationale here might be the fact that personal health information is very sensitive and that we like to keep that 5 information private. When this information becomes public, it might make consumers more prone to social judgment or sanctions (Weiss, Ramakrishna, & Somma, 2006). As a result, when this information is used in personalized advertising, it increases the awareness that this information is public and therefore increases negative feelings (e.g., perceived creepiness of the advertisement) and reduces positive feelings (e.g., perceived relevance of the advertisement).

Second, in the context of governmental communication, there has been little research on the use and the effects of personalization techniques. However, there is some research on the use of personalization by political parties. For example, Boerman and Kruikemeier (2016) found that a promoted tweet from a political party increased skepticism and negatively affected source attitudes, whereas this was not the case when the tweet was posted by a brand. Evidence here indicates that when a political party sends personalized advertisements, it is not welcomed by consumers. Boerman and Kruikemeier (2016) suggest that a political promoted tweet is less obviously advertising and therefore people need to activate their persuasion knowledge. In that case, they will try to react against the promoted tweet,

125 Chapter 5: Going too far? leading to more negative outcomes such as a decrease in perceived trustworthiness of the source. As a result, we conclude that people have learned that due to the personal nature of information in this context (e.g., voting behavior) it is inappropriate to use this type of information in personalized advertising. Therefore, the costs of personalization might outweigh its benefits.

Third, news websites are still experimenting with personalization techniques. News sites mainly use personalization to selectively distribute or show their news stories (Moeller, Trilling, Helberger, Irion, & de Vreese, 2016). On the one hand, using this technique could increase the relevance of news items and thus positive attitudes. On the other hand, it creates the risk for ‘filter bubbles’ which means that consumers do not receive news that is not compliant with their own interests and preferences (Moe &

Schweidel, 2014; Pariser, 2012; Zuiderveen Borgesius et al., 2016). Nevertheless, a survey conducted in 26 countries indicated that, collectively, people prefer algorithmic selection over editorial curation

(Thurman, Moeller, Helberger, & Trilling, 2018). Consumers appear to appreciate personalization algorithms in the context of news websites. Therefore, we conclude that here the benefits might trump the costs, increasing more positive attitudes towards the source of the personalized ads.

Fourth, in a commercial context personalization strategies have been used for a relatively long time now and became common practice (Strycharz, van Noort, Helberger, & Smith, 2019). As a result, consumers encounter personalized advertisements on social networking sites on a daily basis and might have grown accustomed to it. For example, 52% indicate to switch brands when a brand has no personalized communication (Marketo, 2018). Nevertheless, although it appears to trigger positive effects via relevance (e.g. De Keyzer et al., 2015), there is also evidence that it might increase the perception of loss of information control (e.g. De Keyzer et al., 2018; Smit et al., 2014). In sum, in a commercial context, it remains unclear whether the benefits or the costs would have a stronger predictive power.

In sum, we expect that the privacy calculus is context-dependent. Some types of context-related personal information are considered as less sensitive (e.g., previously read articles in news website) compared to others (e.g., personal information in a health context). This is in line with Acquisti et al.’s

(2015) reasoning about context-dependency of privacy behavior: we have learned to keep health information private and as a result, we consider it less appropriate to be used in personalized advertising.

Consequently, we argue that the effects of personalized advertising from different sources may lead to

126 Chapter 5: Going too far? different outcomes. Therefore, we compare the personalization effects and competing mediating processes across four different sources: health, governmental, news and commercial. Would people weigh the costs and benefits from personalization differently according to the source and, accordingly, would the effect of personalization be different across these sources?

RQ2: To what extent do different online sources (i.e., health, governmental, news, and commercial sources) moderate the relationship between personalization and source attitudes, mediated by perceived creepiness and perceived relevance of the advertisement? Method

Study Design and Pretest

To test our conceptual framework (Figure 5.1.) we have set up a 3 (personalization: page-likes from friends vs. page-likes from friends + gender vs. page-likes from friends + gender + interests) x 4

(source: health website, governmental website, online newspaper, and online store) between subject- experiment in which we exposed participants to different personalization scenarios. We exposed respondents to a short, carefully constructed description of a personalized advertisement on Facebook 5 (vignette-based method; Atzmüller & Steiner, 2010). Both in advertising (e.g., Verberckmoes, Poels, Dens, Herrewijn, & De Pelsmacker, 2016) and communication research (e.g., Kruikemeier, van Noort, Vliegenthart, & de Vreese, 2013) the use of vignettes as stimulus materials is very common. It allows for the estimation of unconfounded and context-dependent effects of explanatory vignette factors (Atzmüller & Steiner, 2010). Therefore, the use of vignettes allows for causal investigations of consumer’s responses to personalized advertising (Atzmüller & Steiner, 2010).

Figure 5.1. Conceptual Framework

127 Chapter 5: Going too far?

To select the personalization elements for our main study, in a pretest (N = 45, Mage = 21.24, SDage =

1.52, 15.6% male), each respondent rated the level of perceived personalization of five different vignettes. In total, we created 25 vignettes based on Chapter 2 which examined which cues increase perceived personalization. Respondents were instructed to imagine that they were visiting Facebook where they saw an advertisement that was based on, for example, their gender, interests, and friends’ page-likes. Each vignette contained a maximum of three types of personal data (e.g., their gender, interest and their friends’ page-likes). Respondents were students from a large Dutch university who received credit points for their participation. For a random set of five vignettes participants were asked to rate on a seven-point scale (1 = strongly disagree 7 = strongly agree) to what extent the advertisement was personalized. Based on the findings of one sample t-tests we selected the use of page-likes of friends as the least personalized condition (M = 2.36, SD = .88), which was significantly lower than the mid-point of the scale (t (8) = -3.90, p = .005). Next, the page-likes of friends were combined with the use of the respondents’ gender which was perceived as moderately personalized (M

= 3.18, SD = .89), because this condition was not significantly different from the midpoint of the scale

(t (9) = -1.15, p = .278). Finally, in the third condition the page-likes of friends and gender were complemented by the use of the respondent’s interests (M = 4.85, SD = .96). This combination resulted in a significantly higher perceived personalization than the midpoint of the scale (t (9) = 4.45, p = .002).

Procedure

In the actual study, we focus on Facebook, because it is a popular social network site today in terms of users (Statista, 2018). We created twelve vignettes (3 combinations of personalization elements x 4 online sources) to cover our design. Participants (N = 619, Mage = 45.73, SDage = 15.59, 52.2% male) were recruited [anonymized company, country, and panel for peer review] and randomly assigned to one of the conditions. Only participants with a Facebook account were selected, 65.9% of the respondents spent time on Facebook every day. Only 3.9% of the respondents had less than 11

Facebook friends, and 10.5% of respondents reported having more than 400 friends. Respondents first answered questions regarding their Facebook use: the number of Facebook friends and number of days spent on Facebook in the past week. Before reading the vignette, participants were instructed to imagine the scenario as vividly as possible. All vignettes were similar to the vignettes used in the pretest, except for the use of a specific context which was added in the text by mentioning for example that the

128 Chapter 5: Going too far? advertisement was “for a health website, like gezondheidsplein.nl or thuisarts.nl” (see Appendix 1.E.).

After measuring perceived creepiness, perceived relevance and source attitude participants were debriefed and thanked for their participation.

Table 5.1. Measures Used in the Study Constructs Items Mean Factor Cronbach’s α Composite AVE (SD) Loadings reliability Number of FB How many Facebook friends friends do you have in 189.63 total? (if you are unsure, - - - - (360.51) please make an estimation) Days spent on In the past week, how 5.92 FB many days have you used - - - - 1.78) Facebook? Perceived To what extent do you

creepiness of think the advertisement

the content from the story was: 2.39 .952 .944 .964 .898 1) Creepy (1.62) .943 2) disturbing .949 3) Worrying Perceived What do you think about relevance of the advertisement in the the content story? The advertisement

was:

1) Not important – 2.99 .952 .949 .967 .908 Important (1.63) .952 2) Not relevant – relevant .955 3) Meaningless - 5 Meaningful

Source The (health website), like attitude example 1 and example 2, from the story is: 1) Unattractive – 4.04 .926 .954 .966 .878 Attractive (1.35) .935 2) Bad – Good .953 3) Unpleasant – Pleasant .934 4) Unfavorable - Favorable Measures

Perceived creepiness. Perceived creepiness was measured with three items: “To what extent do you think the advertisement was 1) creepy, 2) disturbing, 3) worrying.” adapted from Zhang and Xu (2016) on a scale from 1 = totally disagree to 7 = totally agree (M = 2.39, SD = 1.42;  = .944).

Perceived relevance. Perceived relevance was measured with three items: “What do you think about the advertisement from the story. The advertisement was: 1) not important – important, 2) not relevant

– relevant, 3) meaningless – meaningful.” from De Keyzer et al. (2015) on a bipolar semantic differential from 1 to 7 (M = 2.99, SD = 1.63;  = .949).

129 Chapter 5: Going too far?

Source attitude. Source attitude was measured with four items: “The (health website/governmental website/online newspaper/commercial website), like (example 1) and (example 2) from the story is 1) unattractive – attractive, 2) bad – good, 3) unpleasant – pleasant, 4) unfavorable – favorable.” on a seven-point semantic differential (M = 4.04, SD = 1.34;  = .954; Spears and Singh (2004).

Descriptive variables. The number of Facebook Friends was measured with one item: “How many

Facebook Friends do you have in total?”. Days spent on Facebook was measured with one item: “In the past week, how many days have you used Facebook?”.

Manipulation checks

The results of the manipulation check indicated significant differences between the different condition in terms of perceived personalization. Post-hoc tests with LSD correction indicate that the condition with page-likes and gender (M = 2.74, SD = 0.94) is not significantly different from the condition with only page-likes (M = 2.87, SD = 1.45, p = .329) in terms of perceived personalization, but it is perceived as significantly less personalized than the condition with page-likes, gender and interests (M = 3.10, SD =

1.48, p = .010). The condition with only page-likes was not significantly different from the condition with page-likes, gender and interests (p = .118).

Results

Confirmatory factor analysis

To examine discriminant validity between our three variables. We first performed a confirmatory factor analysis (CFA) using SmartPLS 3 on the mediating and dependent variables (i.e., perceived creepiness, perceived relevance, and source attitude). Indices of model fit indicate an acceptable fit of the CFA model (SRMR = .034, χ² = 431.388, NFI = .935). After inspecting the factor loadings, we can confirm convergent validity: the factor loadings for all indicators were large and significant (Table 5.1.).

Moreover, the average variance extracted (AVE) for each factor was above the .50 threshold.

Furthermore, reliability estimates range between .964 and .967, which are well over the recommended

.70 (Hair et al., 2010). In Table 5.2., the diagonal shows the square root of the AVE per construct and the off-diagonals show the correlation between each pair of constructs. No correlation was found to be higher than the square root of the AVE, confirming discriminant validity (Fornell & Larcker, 1981).

130 Chapter 5: Going too far?

Moreover, Table 5.2. shows that the average shared variance is larger than the maximum shared variance. Thus, we can confirm discriminant validity (Hair et al., 2010).

Table 5.2. Square root of average variance extracted and correlations per factor

Perceived creepiness of the Positive attitudes toward the Source content content attitude Perceived creepiness of the .937 content Positive attitudes toward the -.306 .948 content Source attitude .551 -.201 .953 Hypotheses testing To test our hypotheses and answer our research questions we used Structural Equation Modeling (SEM) in SmartPLS 3. Indices of model fit indicate an acceptable fit of the structural model (SRMR = .071, χ² = 385.665, NFI = .943). In the first step, the combinations of personalization elements were entered as two independent dummy variables. First, we ran the analysis with the condition including page-likes from friends and gender and the condition including page-likes from friends, gender and interests, using the condition with only page-likes from friends as the reference category. Second, we ran the analysis with the condition with only page-likes from friends and the condition including page-likes from friends, gender and interests, using the condition with page-likes from friends and gender as the reference 5 category.

Perceived creepiness and perceived relevance were included as mediating variables. Source attitude was entered as the dependent variable (Table 5.3.). In a second step, the online source (health, governmental, news, commercial) was entered as a grouping variable for the multi-group comparison (Figure 5.3.).

First, we expected that advertisements with more personalization elements would increase perceived creepiness (H1a), and subsequently result in less positive source attitudes (H1b). The findings show that adding gender to an advertisement with only page-likes from friends does increase perceived creepiness (β = .155, p <.001) and adding both gender and interests to an advertisement with only page-likes from friends increases it even more (β = .209, p < .001). Adding interests to an advertisement with page-likes from friends and gender does not increase the perception of creepiness further (β =

.053, p = .208). H1a is thus partially supported. The findings also confirmed that an increase of perceived creepiness decreases source attitude (β = -.203, p =.<.001). H1b is therefore supported.

131 Chapter 5: Going too far?

Table 5.3. Path coefficients

Reference category: Reference category: Page-likes from Page-likes from friends friends + gender Path p- Path p-value coefficient value coefficient Page-likes from friends  Perceived creepiness of the content - - -.153 <.001 Page-likes from friends  Perceived relevance of the content - - .130 .002 Page-likes from friends + gender Perceived creepiness of .155 <.001 - - the content Page-likes from friends + gender Perceived relevance of the -.133 .004 - - content Page-likes from friends + gender + interests Perceived .209 <.001 .056 .208 creepiness of the content Page-likes from friends + gender + interests Perceived -.055 .245 .076 .091 relevance of the content Perceived creepiness of the content  Source attitude -.203 <.001 -.203 <.001 Perceived relevance of the content  Source attitude .511 <.001 .511 <.001

This is also demonstrated in Figure 5.2. For the indirect effects we find a negative, significant indirect effect from personalization through perceived creepiness on source attitude when going from an advertisement with only page-likes from friends to an advertisement with page-likes from friends and gender (β = -.032, CI = [-.054; -.014], p = .002) and when going to an advertisement with page-likes from friends, gender and interests (β = -.043, CI = [-.068; -.022], p = .001). These findings indicate that the tipping point for creepiness already lies at the addition of gender to an advertisement with only page-likes, adding interests as a personalization element does not increase feelings of creepiness further. Advertisements with gender and interests are perceived as creepier, resulting in more negative source attitudes.

Second, we expected that adding personalization elements would increase perceived relevance (H2a), and subsequently result in more positive source attitudes (H2b). Contrary to what was expected, adding gender to an advertisement with page-likes decreased perceived relevance (β = -.133, p = .004) and adding both gender and interests does not significantly influence perceived relevance (β = -.055, p =

.235). An advertisement with page-likes from friends and gender and an advertisement with page-likes from friends, gender and interests do not differ in perceived relevance (β = .076, p = .091). Therefore,

H2a is not confirmed. However, our results support H2b: a higher level of perceived relevance leads to a more positive source attitude (β = .511, p <.001). This is also demonstrated in the indirect effects of personalization through perceived relevance: we only found one significant indirect effect. An advertisement with page-likes from friends and gender has a negative, significant indirect effect (β = -

.068, CI = [-.114; -.020], p = .003) compared to an advertisement with only page-likes from friends,

132 Chapter 5: Going too far? but not for an advertisement with page-likes from friends, gender and interests compared to an advertisement with only page-likes from friends(β = -.028, CI = [-.072; .018], p = .236) nor compared to an advertisement with page-likes from friends and gender(β = .039, CI = [-.007; -.0083], p = .108). This means that our findings do not support the notion that perceived relevance is a positive explanatory mechanism of personalization effects on source attitude.

For our first research question, we wondered which mediating variable was stronger in explaining the effects of personalization elements to source attitude. We can be short in answering this question because perceived relevance did not positively explain personalization effects. This means that perceived creepiness (compared to perceived relevance) was a negative and stronger predictor of personalization for all conditions.

Figure 5.2. Bootstrap intervals of specific indirect effects of the personalization elements

Notes: The boxes show respectively 1) ULCI, 2) path coefficient and 3) LLCI. Significant indirect effects in bold.

For our second research question, we wondered to what extent the different contexts moderate the relationship between personalization and source attitude, mediated by perceived creepiness and

133 Chapter 5: Going too far? perceived relevance of the advertisement. Therefore, in a second step, we entered the online source type as a grouping variable to examine whether the explanatory power of the mediating variables differs between the various online source types (see Figure 5.3.). When considering context, the previously discussed patterns for perceived creepiness (as a negative explanatory mechanism) and perceived relevance (as a non-significant mediating principle) remain the same, but mainly when comparing the advertisement with page-likes from friends, gender and interests and the advertisement with only page- likes from friends (and not comparing the advertisement with page-likes from friends with the advertisement with page-likes from friends and gender). In the column ‘Page-likes + gender + interests vs page-likes’ column in Figure 5.3. we clearly see that perceived creepiness is a negative explanatory mechanism for a health, a governmental website, and an online newspaper. For an online newspaper, this pattern was also true when comparing the advertisement with only page-likes and the advertisement with page-likes from friends and gender. Interestingly, perceived creepiness does not explain personalization effects in the commercial context (i.e., online store).

134

135 Figure 5.3. Bootstrap intervals of specific indirect effects of personalization for each source type far? too Going 5: Chapter

Notes: The boxes show respectively 1) ULCI, 2) path coefficient and 3) LLCI. Significant indirect effects in bold.

Chapter 5: Going too far?

Discussion

Previous research indicates that people are unaware of which personal data is collected about them or how it is collected (Bucher, 2017). Even more, people do not understand how this personal data is used by algorithms to personalize advertising on, for example, social networking sites. Nevertheless, when using social networking sites, users are confronted with personalized advertisements. The current study examines the implications of that, by investigating the tipping point of personalization: when do advertisers go too far in personalizing the advertisements they provide? And whether this ultimately has a negative effect on the source of the ads. We tested the effects of three different personalized advertisements on source attitude to examine this tipping point. Using the privacy calculus model (e.g.

Culnan & Armstrong, 1999; Dienlin & Metzger, 2016; Laufer & Wolfe, 1977), the current study aimed to examine how two competing mechanisms (i.e., perceived creepiness and perceived relevance) mediate that effect. Furthermore, based on the notion of context-dependency of Acquisti et al. (2015) we proposed and tested how such effects differ across four online sources reflecting different contexts

(health website, governmental website, online news website, and an online store).

In general, our findings suggest that adding gender and interests to advertisement with only page-likes from friends does indeed increase creepiness. More specifically, our results indicate that the sweet-spot, or tipping point, of personalization is found from advertisements with only page-likes from friends.

Adding gender to an advertisement with page-likes from friends as a personalization element increases perceived creepiness but adding interests to that does not increase it further. This might result from the fact that social network site users might not be aware of the fact that so much information is gathered about them and can be used in advertising. As a result, when confronted with the use of this personal information in an advertisement, they might feel as if someone has been ‘snooping’ around in their personal profile or their surfing behavior, leading to perceptions of creepiness, which is line with, for example, Malheiros et al. (2012) and Ur et al. (2012). This increase in creepiness leads to less positive attitudes toward the source. This is in line with what White et al. (2008) have called ‘personalization reactance’. When consumers feel threatened, for example by someone or something examining their personal information, they will react to this threat by behaving opposite than what is intended by the threat. Contrary to our expectations, perceived relevance does not significantly change when going

136 Chapter 5: Going too far? adding interests to an advertisement with only page-likes from friends or to an advertisement with page- likes from friends and gender , but it does, contrary to our expectations, decrease when going adding gender to an advertisement with only page-likes from friends. This might have to do with the fact that people are accustomed to highly personalized messages on social networking sites, as they are confronted with them on a daily basis. Moreover, previous research already indicated that it is important for personalization effects to take place that consumers perceive the actual personalization as personalized (Kramer, 2007; Li, 2016). Even though we based the choice for the personalization elements on the findings of Chapter 2 and on a pretest, incrementally increasing the number of personalization elements did not automatically lead to an incremental increase in perceived personalization. Although an increase in perceived relevance increased the attitude toward the source, we cannot conclude that perceived relevance positively explained personalization effects on source attitudes. This means that perceived creepiness was a stronger explanation for personalization effects.

Finally, we examined whether different sources, reflecting distinctive contexts, moderated the effects on source attitude via perceived creepiness and perceived relevance. Our findings clearly suggest that there are indeed differences between types of advertisers. The negative impact of personalization through perceived creepiness holds for health and governmental sources and for an online newspaper. 5 Interestingly, this pattern does not occur within a commercial context, that is for online stores. In a commercial context, creepiness does not seem to play a role. Consumers might have grown accustomed to receiving highly personalized messages of commercial sources and, therefore, might not be creeped out by them (De Keyzer et al., 2018). In sum, this means that personalization effects and the privacy- calculus theory are context-dependent.

Theoretical and Practical Implications

Our results have several important implications. A first theoretical implication is that our findings suggest that the privacy calculus does not only apply to self-disclosure behavior (Demmers et al., 2018) but also to other important online consumer responses, that is the evaluation of the message source. There appears to be some relation between how consumers believe they should behave (self-disclosure) and how they evaluate sources of personalized advertising. Our findings suggest that in the context of social networking sites the benefits of personalized advertising do not seem to trump the costs of personalized

137 Chapter 5: Going too far? advertising. This means that even though previous research has concluded that consumers seem to appreciate personalized advertisements (Ur et al., 2012), our findings suggest that they are primarily creeped out. More specifically, we looked for the sweet spot of personalized advertising: when does personalization come too close? Our findings indicate that, in general, this is already when only using page-likes from friends: adding the personalization elements gender and interests increases creepiness while it does not increase perceived relevance. This means that using only page-likes from friends seems to creep out consumers the least and adding gender already significantly increases creepiness. In line with Malheiros et al. (2012) and Ur et al. (2012), we found that perceived creepiness negatively affects source attitudes. This indicates that when personalization goes too far for a consumer, (s)he will evaluate the source as less positive. Finally, we examined the moderating effect of source type. In line with Acquisti et al. (2015) and Nissenbaum (2010), we find that the tipping point of personalization is context-dependent. The negative impact of personalization through perceived creepiness only occurred for health websites, governmental websites and online newspapers, but not in a commercial context.

This indicates the way in which people weigh the benefits and risks of personalized advertisements differently in specific situations. Especially in situations where the stakes are higher and people might less likely be accustomed to personalized advertising, people weigh the risks as more important.

Our findings also have far-reaching practical implications. On the one hand, our findings indicate that personalized advertising on SNSs already appears to be perceived as creepy by consumers when it is moderately personalized (using page-likes from friends and gender), leading to less positive consumer responses. On the other hand, relevance did not increase with the number of personalization elements.

In general, it is, therefore, best to use only a small number of personalization elements. We would suggest to not combine several different personalization elements in one advertising. Moreover, Chapter

2 indicates that interests, location and age are best in eliciting perceptions of personalization. When using one of these personalization elements, we would advise advertisers not to combine these personalization elements with other elements. Examining the effects separately for the four different source types helped to clarify these effects. In a commercial context, adding gender and interestsdoes not matter, as advertisments containing these personalization elements are not perceived as creepier than an advertisement with only page-likes from friends. However, for other contexts (health, governmental and news) one should be careful with the usage of personal data for personalization.

138 Chapter 5: Going too far?

Especially the use of a higher number of personalization elements is perceived as creepy and negatively impacts how consumers evaluate the source. This means that in these contexts, personalization might negatively impact the brand. As brand responses are strongly related to all kinds of consumer responses - also responses that are specifically important in online and social media platforms (i.e., engagement, reference) - personalization should be handled with care. Commercial brands can take it a lot further, although we are not able to demonstrate any positive effects, through relevance, of using more personalization elements. This means that commercial brands possibly do not monetize on personalization efforts in social media advertising.

Agenda for Future Research The limitations and findings of this study provide opportunities for future research. In the current study, respondents were only exposed to a vignette asking them to imagine a certain situation in which they encountered a personalized advertisement. Although this allows for an internally valid experiment, it also reduces external validity.

Moreover, our study only used self-reported attitudes and did not measure actual behavior (e.g., click- through rates, engagement responses, or sales). Current data analyzing techniques allow researchers to analyze real social network data and its impact on for example click-through rates and might be a 5 venue for future research. Nevertheless, self-reported measures are common in experimental research for reasons of internal validity or when people are confronted with fictitious brands (Geuens & De Pelsmacker, 2017). Also, when analyzing real social network data, it becomes difficult to compare different personalized advertising messages and its effects because they might differ on more aspect than merely their level of personalization and the source. Therefore, internal validity would be compromised.

We now have used only three different personalization elements: page-likes from friends, gender and interests. These are based on Chapter 2 and on a pretest. However, to conclude where exactly the tipping point of personalization lies, we suggest further research to include more different personalization elements in different combinations to further specify the tipping point of personalization.

Moreover, future research should examine how different personalization elements contribute to perceptions of personalization for different types of sources. In Chapter 2 we used different products

139 Chapter 5: Going too far? and services to test which personalization elements would be most important to elicit perceived personalization. Nevertheless, they were all commercial. The current study indicates that the use of the personalization elements page-likes from friends, gender and interests lead to different personalization effects for different sources. As such, it might be that for different types of sources different personalization elements are more successful in eliciting perceived personalization.

After exposure to the stimulus, perceived creepiness was measured before perceived relevance or source attitude. The order of the measures in the survey might have affected the results. Future research could test the robustness of our findings by changing the order of the measures. Finally, future research could also look into other potential moderating variables such as experience with personalized advertising or knowledge about personalized advertising. Based on Persuasion Knowledge Model (PKM) from Friestad and Wright (1994) it could be expected that as consumers learn about personalization tactics the effect of its creepiness decreases. After all, creepiness has to do with the fact that consumers are uncertainty and unpredictability of a certain behavior, i.c., personalized advertising.

Finally, in line with the previous chapters, we find that the direct effects of perceived relevance (β =

.510, p < .001) and perceived creepiness (β = -.203, p <.001) on source attitude are stronger than the direct effect of personalization (βpage-likes from friends + gender = -.098, p = .036; βpage-likes from friends + gender + interests

= -.068, p = .147), indicating that other factors contribute to the perceptions of relevance and creepiness.

In conclusion, our findings contribute to the research on personalized advertising on social networking sites. More specifically, it examined the cost and benefit side of personalization based on the privacy calculus theory. We find evidence for the fact that the benefits seem to outweigh the costs. Furthermore, we examined the moderating role of source type. It becomes clear that this helps to explain the fact that previous studies found both positive as well as negative effects of personalization (e.g. De Keyzer et al., 2015; van Doorn & Hoekstra, 2013). It matters in which context personalization takes place and future endeavors focusing on the implications of personalization should take that into account.

140

PART II - Word-of-Mouth on Social Networking Sites-

If you make customers unhappy in the physical world, they might each tell six friends. If you make

customers unhappy on the Internet, they can each tell 6,000.

-- Jeff Bezos, Amazon

Don’t be so emotional! How tone of voice and service type affect the relationship between message valence and consumer responses to WOM in social media1, 2

1 Manuscript published as De Keyzer, F., Dens, N., & De Pelsmacker, P. (2017). Don't be so emotional! How tone of voice and service type affect the relationship between message valence and consumer responses to WOM in social media. Online Information Review, 41(7), 905-920. 2 An earlier version of this chapter was presented at 2016 International Conference on Research in Advertising, Ljubljana, Slovenia. 6 Chapter 6: Don’t be so emotional!

Abstract

The purpose of this article is to shed light on the boundary conditions of the effect of the valence of word-of-mouth on social networking sites (sWOM) on consumer responses (attitude toward the service provider, purchase intention and positive word-of-mouth intention). Specifically, we examine two moderators: the tone of voice (factual versus emotional) of the sWOM and service type (utilitarian versus hedonic) of the service that the sWOM is about. A 2 (message valence: positive versus negative) x 2 (tone of voice: factual versus emotional) x 2 (service type: utilitarian versus hedonic) full-factorial between-subjects online experiment with 400 respondents was conducted and the data were analyzed using Hayes’ PROCESS macro. The results show that message valence exerts a greater impact on consumer responses with factual sWOM messages compared to emotional ones. Furthermore, the impact of message valence is stronger for more hedonic services compared to more utilitarian services.

In contrast to our expectations, there is no significant impact of matching the tone of voice to the service type. First, for sWOM senders, factual messages are found to be more influential: backing an sWOM up with arguments and specific details increases the chance of it affecting consumers’ responses. As a result, marketers, especially of predominantly hedonic services, should encourage their followers and customers to spread positive factual sWOM about their service. The study tests two previously unstudied moderating variables that affect the relationship between message valence and consumer responses to sWOM messages. Moreover, this study provides interesting insights for marketers and bloggers or reviewers.

144 Chapter 6: Don’t be so emotional!

Introduction

Consumers’ increased use of online communications is reflected in their word-of-mouth (WOM) behavior

(Karjaluoto, Munnukka, & Kiuru, 2016). Sharing daily consumption online is an important part of modern life (Kim, Jang, et al., 2015). Hennig-Thurau et al. (2004) define electronic word-of-mouth (eWOM) as:

“any positive or negative statement made by potential, actual, or former customers about a product or a company, which is made available to a multitude of people and institutions via the Internet.” (p. 39). eWOM can occur through review sites (e.g., TripAdvisor), (micro) blogging platforms (e.g., Twitter), video sharing sites (e.g., YouTube) and social networking sites (SNSs, e.g., Facebook). In the present study, we are interested in the latter, for which the term sWOM has been coined (Balaji et al., 2016). sWOM is eWOM on SNSs, which are, “applications that enable users to connect by creating personal information profiles, inviting friends and colleagues to have access to those profiles, and sending e- mails and instant messages between each other” (Kaplan & Haenlein, 2010, p. 63).

In general, eWOM is acknowledged to greatly impact consumer attitudes, behavioral intentions (e.g. Cheung & Thadani, 2012) and sales (Rui et al., 2013). Sociable Labs (2012) report that 62 percent of online shoppers have read product-related messages from Facebook connections. The limited research on sWOM seems to suggest that sWOM, like other forms of eWOM, is perceived by consumers as 6 trustworthy and credible (Chu & Kim, 2011) and exerts a positive effect on consumers’ online brand engagement, brand awareness, brand attitude and purchase intention (PI) (Karakaya & Barners, 2010; Schivinski & Dabrowski, 2015; Wang et al., 2012).

It may, however, be important to distinguish sWOM from the broader concept of eWOM. First, research on eWOM already suggests that not all digital WOM is created equal. In an explicit comparison between consumer reviews and microblogs, Marchand, Hennig-Thurau, and Wiertz (2017) document differential effects of the valence and volume of the two types of eWOM on video game sales. Second, some research documents differences between eWOM and sWOM (e.g. Lin et al., 2017). From a sender’s perspective, one of the major implications of sWOM is that the sender is identifiable and can be held accountable, which results in a higher social risk (Balaji et al., 2016; Eisingerich et al., 2015), while other forms of eWOM (e.g., online reviews) may be anonymous. From a reader’s perspective, sWOM has a unique advantage of referability: the contributor’s SNS profile provides for a higher level of source

145 Chapter 6: Don’t be so emotional! credibility, benefiting sWOM adoption (Hajli, 2016). Identification with the sender is also a relatively unique aspect of SNSs that drives readers’ PI (Wang et al., 2012). Bachleda and Berrada-Fathi (2016) suggest that negative sWOM from a Facebook friend is less influential than negative eWOM from a consumer review site because readers place less trust in sWOM.

Users’ motivation to visit SNSs and review sites also differs, which could result in different responses to

WOM on these platforms (Gvili & Levy, 2016). SNSs are mainly used to pass time and for amusement and social exchange (Ku et al., 2013), whereas review sites are consulted to read information on a product or service in which the user is already interested (Reichelt, Sievert, & Jacob, 2014). Exposure to brand information in sWOM is more voluntary (Chu & Kim, 2011). The meta-analysis of Babić Rosario,

Sotgiu, De Valck, and Bijmolt (2016) indicates that the effect of sWOM on sales is weaker than the effect of eWOM through e-commerce platforms. sWOM does, on the other hand, entail an increased risk that customer complaints go viral (i.e., shared on a massive scale on SNSs), causing a potential public relations crisis for a firm. It is therefore essential that the determinants and consequences of sWOM are examined (Balaji et al., 2016).

The beneficial effect of positive message valence on attitudes (Purnawirawan et al., 2015), brand loyalty and perceived brand quality (Schivinski & Dabrowski, 2015), PIs (Bae & Lee, 2011) and even sales

(Floyd et al., 2014) of the message recipients is consistently found in the eWOM literature (e.g. Floyd et al., 2014; Lee & Koo, 2012; Purnawirawan et al., 2015), and has been replicated in one sWOM study

(Rui et al., 2013). However, this effect can be reinforced or weakened by moderating factors.

The current study offers a number of contributions. According to Barger et al.’s (2016) conceptual framework on consumer engagement in social media, an antecedent that needs more extensive research is the “content factor.” Our study responds to this call by inspecting how the tone of voice of a message

(whether the sWOM message takes on a more factual or emotional tone) moderates the effect of message valence on consumer responses. Moreover, we also inspect the moderating effect of service type (whether the service mainly fulfills utilitarian or hedonic buying motivations) on the effect of sWOM valence, which in the framework proposed by Barger et al. (2016), can be categorized as a “product factor.” By studying the combination of these variables, we are able to explore the boundary conditions of the effects of sWOM message valence.

146 Chapter 6: Don’t be so emotional!

This study sets out to corroborate findings on the role of message valence and tone of voice previously found in an eWOM context in an sWOM, context and further tests the moderating effect of service type. We test our propositions for services since sWOM has been found to be more influential when services are discussed (compared to products) (Babić Rosario et al., 2016). We focus on both traditional brand responses (attitude and PI) (e.g. Park & Lee, 2008; Purnawirawan et al., 2015), and consumer engagement, more specifically, positive WOM intention. Consumer engagement is “a consumer’s positively valenced brand-related cognitive, emotional and behavioral activity during or related to focal consumer/brand interactions” (Hollebeek, Glynn, & Brodie, 2014, p. 154). Positive WOM intention is a particularly relevant outcome variable in the context of social media, and SNSs in particular (Geuens & De Pelsmacker, 2017).

Literature review and hypotheses

The influence of message valence on consumer responses to sWOM Most studies find that positive eWOM benefits recipients’ attitudes (Doh & Hwang, 2009; Purnawirawan et al., 2015), PIs (Bae & Lee, 2011; Doh & Hwang, 2009) and sales (Floyd et al., 2014), while negative eWOM entails negative effects. In an sWOM context, Rui et al. (2013) report a positive effect of positive tweets about a movie on movie sales and a negative effect of negative tweets. While the effect of 6 message valence is quite consistent, it is influenced by a number of moderators. For example, the effect of review valence is greater when consumers have less product expertise or are not familiar with the brand (Purnawirawan et al., 2015; Zou, Yu, & Hao, 2011). López-López and Parra (2016) show that the presence of a review voted as “the most helpful” influences recipients’ attitude toward the product in the direction of the review valence. Moreover, they find that this effect is further reinforced by goal congruency between the focal review and the recipients’ goals.

Inspired by Barger et al. (2016), we examine how message content and product-related antecedents, such as a factual vs emotional tone of voice and the service type moderate the effects of sWOM message valence. The conceptual model on which the hypotheses in the following sections are based, is presented in Figure 6.1.

147 Chapter 6: Don’t be so emotional!

Figure 6.1. Conceptual framework

Tone of voice

Service type H3 H4 H1

a) Attitude toward the service provider Message valence b) Purchase intention c) WOM intention

The moderating effect of a factual vs an emotional tone of voice

We consider the factual vs emotional tone of voice of sWOM as an element of message content, an important antecedent of consumer engagement on social media (Barger et al., 2016). Some messages are (predominantly) factual, based on attribute-value information, such as “My internet connection speed is twice as fast as before.” The arguments used are rational, objective, specific and clear. Other messages are predominantly emotional, focusing on the feelings caused by the experience of using the product by the writer, with no or little support from verifiable arguments (e.g., “This hotel was awesome!

I’m really glad we stayed here”) (Park & Lee, 2008). Emotional messages are often more subjective, and abstract, containing interjection and non-relative information.

Previous research uses different terminology to examine relatively similar phenomena. Huang, Chou, and Lan (2007) find that negative online reviews lead to a higher opinion acceptance, WOM intentions, and boycott intentions when the reviews are instrumental (cf. factual) compared to affective (cf. emotional). (Park & Lee, 2009) show that positive attribute-value reviews (cf. factual) that focus on the attributes of a portable multimedia player increase PI more than recommendations that focus on the experience of using the product (cf. emotional). Wu and Wang (2011) conclude that positive rational eWOM messages outperformed positive emotional ones with respect to their effect on brand attitude, trust and affection and PI, especially for highly involved consumers. Lee and Koo’s (2012) findings suggest that objective (cf. factual) information has a stronger effect on consumer attitudes and behavior

(intentions) than subjective (cf. emotional) information. They suggest that this results from the fact that objective information will less likely result in misunderstanding or bias in the evaluation process. Chun and Lee (2016) confirm these findings in an sWOM context in that users perceive messages with more

148 Chapter 6: Don’t be so emotional! utilitarian-value (cf. factual) content as more useful than hedonic-value (cf. emotional) messages. This, in turn, increases users’ behavioral intention to subscribe to a company’s Facebook page, learn about the company, recommend the use of the company and promote the company. However, they considered only positive messages and did therefore not study the moderating effect of the tone of voice on the effect of valence.

These results can be explained by the accessibility-diagnosticity theory (Ahluwalia & Gürhan-Canli, 2000). Information needs to be both accessible in consumers’ memory and diagnostic before it influences evaluations. Consumers evaluate the diagnosticity of information by its ability to help them in evaluating the quality and performance of the target object. Diagnostic information that is less ambiguous will more likely be used. Factual messages, based on concrete product characteristics, independent of the reviewer, are perceived as more informative or diagnostic. This might be especially important for services as they cannot be seen or touched (Sweeney, Soutar, & Mazzarol, 2012).

The findings are also explained by attribution theory (Moran & Muzellec, 2014; Sen & Lerman, 2007). A message can be attributed to product performance (stimulus attribution) and/or to dispositional characteristics of the communicator (non-stimulus attribution) (Lee & Youn, 2009; Moran & Muzellec, 2014). Factual messages are more likely to induce stimulus attribution since the use of arguments 6 provides a better insight into product performance. In contrast, emotional messages are more likely to induce non-stimulus attribution due to the lack of arguments and the use of emotional statements.

Subsequently, the information will be discounted in the evaluation of the product’s actual performance

(Lee & Youn, 2009; Moran & Muzellec, 2014).

In line with these theories and results, we expect that factual messages will reinforce the effect of message valence. Positive sWOM should obviously result in more positive consumer responses than negative ones, but the difference in effects between positive and negative sWOM should be greater with factual than with emotional messages:

H1: The effect of sWOM message valence on readers’ (a) attitude toward the service provider (b) PI and (c) WOM intention is stronger for sWOM messages that are perceived as more factual.

149 Chapter 6: Don’t be so emotional!

The moderating effect of service type

People use goods and services based on hedonic or utilitarian motivations (Batra & Ahtola, 1991; Dhar

& Wertenbroch, 2000). While the consumption of most goods and services can involve both hedonic and utilitarian dimensions, researchers often distinguish between predominantly utilitarian and hedonic products (Batra & Ahtola, 1991; Dhar & Wertenbroch, 2000; Hellén & Sääksjärvi, 2011). When buying a service for predominantly hedonic motivations (e.g., restaurant, bar), people seek value from an affective and sensory experience of aesthetics, sensual pleasure, fantasy and fun (Hirschman &

Holbrook, 1982). The experience and derived sensations from consuming the service drive quality perception and satisfaction. Consequently, hedonic services are highly person-specific and are mostly experienced subjectively (Voss et al., 2003). A service bought for utilitarian motivations (e.g., cell phone provider, bank), on the other hand, is used to accomplish a functional or practical task. As these tasks are less person-specific, an evaluation of the service quality and customer satisfaction can be made more objectively (Voss et al., 2003). Hedonic and utilitarian buying motivations lead to differences in information processing (King & Balasubramanian, 1994; Senecal & Nantel, 2004). Due to the fact that hedonic services cannot be known or evaluated until experienced or used (Klein, 1998; Zhu & Zhang,

2010), pre-purchase uncertainty is high (Park & Lee, 2009; Park & Park, 2013). Risk aversion causes consumers to perceive eWOM about a hedonic service as more diagnostic (Willemsen et al., 2011).

Utilitarian services, on the other hand, can be evaluated based on tangible attributes prior to purchase

(Klein, 1998; Purnawirawan et al., 2015). As a result, consumers may complement sWOM with other ( factual) information that they can readily observe. Therefore, sWOM will be relatively less diagnostic for utilitarian products and services than for hedonic ones (Willemsen et al., 2011). Based on accessibility- diagnosticity theory, message valence should, therefore, exert a greater effect on readers’ responses for hedonic than for utilitarian products. Previous research in this field has focused exclusively on online reviews. However, we assume that these findings also apply in an sWOM context. Therefore, we expect:

H2: The effect of sWOM message valence on readers’ (a) attitude toward the service provider (b) purchase intention (c) WOM intention is stronger for a hedonic service than for a utilitarian service.

150 Chapter 6: Don’t be so emotional!

Tone of voice, service type and consumer responses

In the previous sections, we developed hypotheses on how the message tone of voice and the service type moderate the effect of valence separately. In this section, we will hypothesize that, additionally, all three factors interact to influence consumer responses. Based on the matching principle (e.g. Klein

& Melnyk, 2016), we expect that a match between the message tone of voice and the service type exerts a greater impact on consumers’ responses. The matching principle states that advertising messages that are compatible with a consumer’s motivations (utilitarian or hedonic) are more persuasive than incompatible messages (Shavitt, 1990). López-López and Parra (2016) indeed found that the effect of a review voted as the most helpful is stronger when the review is congruent with the reader’s goals.

The consumption of utilitarian goods is more cognitively driven and goal-oriented toward a functional or practical task. Therefore, consumers tend to evaluate utilitarian services more cognitively. When processing information for this type of products, the immediate consequences of using the products will be of interest (Batra & Ahtola, 1991). As hypothesized in H1, the use of tangible and objective information in WOM (i.e., a factual tone of voice) will increase the likelihood that readers rely on the

WOM (Grabner-Kräuter & Waiguny, 2015). The matching principle prescribes that a more factual tone of voice is especially influential when the sWOM concerns a utilitarian service, as a factual tone of voice 6 matches with readers’ utilitarian buying motivations. Therefore, we expect:

H3: For sWOM messages about a utilitarian service, the effect of message valence on readers’ (a) attitude toward the brand (b) PI (c) WOM intention is stronger when the message is perceived as more factual.

Hedonic services satisfy emotional needs. As a result, consumers perceive affect and experiences as important when evaluating a hedonic service (Hellén & Sääksjärvi, 2011). Emotional messages will, therefore, be more diagnostic (Jiang & Wang, 2006). Research shows that consumers use more signs of affect when evaluating hedonic services compared to utilitarian services (Jiang & Wang, 2006) and eWOM about these goods and services are more affectively processed (Klein & Melnyk, 2016). Thus, sWOM messages using an emotional tone of voice might match better with hedonic services (Grabner-

Kräuter & Waiguny, 2015). We expect:

151 Chapter 6: Don’t be so emotional!

H4: For sWOM messages about a hedonic service, the effect of message valence on readers’ (a) attitude toward the brand (b) PI (c) WOM intention is stronger when the message is perceived as more emotional.

Study design and procedure

We created a 2 (valence: positive vs negative) x 2 (tone of voice: emotional vs factual) x 2 (service type: utilitarian vs hedonic) full-factorial between-subjects experiment. Although the main study is conducted with U.S. participants, for convenience reasons two pretests were conducted with Belgian participants. The scales used in the pretests are the same as in the main study (Table 6.1.). In a first pretest (N=36, Mage=46.81, 38.9% male), respondents rated how utilitarian and hedonic a number of services are. We also measured involvement as that is a potential confound in responses to emotional and factual messages. Respondents were recruited via Facebook to complete an online questionnaire.

We selected a bar as a hedonic service, and a cell phone provider as a utilitarian service, as these differed significantly in perceived hedonism ( Mbar=5.96, Mprovider=4.48, p<.001)) and perceived utilitarianism ( Mprovider=6.23, Mbar=4.86, p<.001), but not in involvement (p= .369).

Next, we created eight potential sWOM messages to cover our 2x2x2 design (see Appendix A.E). All messages consisted of two short sentences followed by two hashtags. The emotional messages reflect the emotions the sender experienced in using the service (e.g., “It is pure enjoyment!”). The factual messages provide service-relevant information and factual characteristics (e.g., “It was also clean as a whistle.”), which is in line with the manipulations of Huang et al. (2007) and Park and Lee (2008). In a second pretest, 166 respondents (age range: 19 to 61, Mage= 38.40 (SD= 12.66); 31.9% male) were recruited by email from an online panel set up by the researchers’ department to complete an online questionnaire. Each respondent was randomly assigned to a single valence by tone of voice condition

(e.g., positive valence – factual tone). That is why the number of respondents is higher than in the first pretest. Respondents rated the perceived valence (negative/positive) and perceived tone of voice (e.g., not factual/factual) for the created sWOM message in their condition for both the bar and the cell phone provider (within-subjects), in a randomized order. An independent samples t-test indicated that the factual messages were indeed perceived as more factual than the emotional ones and the positive

152 Chapter 6: Don’t be so emotional! messages were indeed perceived as significantly more positive than the negative messages (all p<

.001).

The main study was set on Facebook because Facebook is considered the most popular social network site today (cf. Fang, 2014; Wang & Chang, 2013). We drafted a (non-interactive) mock Facebook page to mimic an actual Facebook news feed, in line with previous research (Chun & Lee, 2016). We instructed respondents to imagine that this was their own news feed. In the feed (among other things, which were held constant across conditions), a (fictitious) Facebook friend had posted a message about a bar (hedonic service) or a cell phone provider (utilitarian service). Both services were fictitious brands to avoid potential confounds of prior brand knowledge and attitudes. The post was either positive or negative and written either in an emotional or factual tone of voice. American respondents (N=400,

Mage=28.15, 54.5% male) were recruited for an online survey via a professional recruitment service and randomly assigned to conditions. All respondents were at least undergraduates. Only participants with a Facebook account were selected, 50.5% of the respondents spent time on Facebook every day, with a majority (56.3%) spending between 10 and 30 minutes per day on Facebook. Only 4.3% of respondents had less than 11 Facebook friends and 18.5% had more than 401 Facebook friends. The average score on a Facebook intensity scale was 4.34 (SD= 1.67). 6 Measures

Table 6.1. shows all construct items. Respondents first answered the questions on their Facebook use,

Facebook intensity (e.g., “Facebook is part of my everyday activity”), number of days spent on Facebook in the past week, average minutes per day spent on Facebook and number of Facebook friends.

Respondents then saw the mock Facebook page containing the target sWOM post, with the instruction to imagine that they were looking at their own Facebook news feed. Next, participants rated the perceived valence and tone of voice of the target post, their attitude toward the service provider (Att),

PI and (positive) WOM intention. Finally, respondents were asked to rate the hedonic and utilitarian buying motivation for either the cell phone or the bar. All constructs were measured by means of seven- point Likert scales or semantic differentials. Construct scores were computed by calculating the average of the items per construct.

153 Chapter 6: Don’t be so emotional!

Table 6.1. Construct items

Construct Items Cronbach’s Scale Origin Alpha in the Main Study Perceived message To what extent do you think the post was -

valence negative or positive? - Emotional – rational .735 - Intangible - tangible Perceived tone of Liu and Stout - Subjective - objective voice (1987) - Nonfactual - factual - Nonlogical - logical - Not fun – fun .978 Hedonic buying - Dull - exciting Voss et al. (2003) motivation - Unenjoyable - enjoyable - Ineffective – effective .927 Utilitarian buying - Not functional - functional Voss et al. (2003) motivation - Impractical – Practical - Unappealing – appealing .918 Attitude toward - Bad - good Spears and Singh service provider - Unpleasant – pleasant (2004) - Unfavorable - favorable If I were to choose a bar/cell phone provider, I - Dodds, Monroe, and Purchase intention would consider Chromebar/Smartline. Grewal (1991) - I am likely to say negative/positive things .842 about Chromebar/Smartline. Word-of-mouth - I am (not) likely to recommend Brüggen, Foubert, intention Chromebar/Smartline to a friend or a colleague and Gremler (2011) - I am likely to discourage/encourage friends and relatives to visit Chromebar/use Smartline. - Facebook is part of my everyday activity. .838 - I feel out of touch when I haven’t logged FB Intensity Ellison et al. (2007) onto Facebook for a while. - I would feel sorry if Facebook shut down. - In the past week, on average, approximately - Days spent on FB Ellison et al. (2007) how many days have you used Facebook? Minutes per day - In the past week, on average, how many - Ellison et al. (2007) spent on FB minutes per day have you spent on Facebook? Number of FB - About how many Facebook friends do you - Ellison et al. (2007) friends have in total? A cell phone provider/bar is … to me. .947 Product category Unimportant – Important De Meulenaer et al. involvement Meaningless – Meaningful (2015) Does not matter to me – Does matter to me

Manipulation checks

Although the pretests were conducted in a different country than the main study, in line with the results of our pretests, all manipulations were successful. Participants rated the positive sWOM (M= 5.94, SD=

1.13) as significantly more positive than the negative sWOM (M= 1.84, SD= .91, t(398)= -40.12, p<

.001). The factual sWOM (M= 4.57, SD= 1.11) was rated more factual than the emotional sWOM (M=

3.85, SD= 1.10, t(398)= 6.51, p< .001). Next, a cell phone provider (M= 6.35, SD= .86) was rated as more utilitarian than a bar (M= 4.95, SD= 1.20, t(398)= 6.51, p< .001), while a bar (M= 5.75, SD=

154 Chapter 6: Don’t be so emotional!

.1.21) was rated as more hedonic than a cell phone provider (M= 4.77, SD= 1.10, t(398)= 8.88, p<

.001).

Results Appendix 3.D. shows the results from the analysis with the actual manipulations. The results indicate that message valence and service type both have a positive effect on attitude toward the service provider (Bmessage valence = 1.807, p < .001; Bservice type = .479, p < .001), purchase intention (Bmessage valence

= 1.442, p < .001: Bservice type = .790, p < .001) and WOM intention (Bmessage valence = 880, p < .001;

Bservice type = .464, p < .001). However, tone of voice does not (Battitude toward the service provider = .119, p =

.286; Bpurchase intention = .212, p =.106, BWOM intention = -.110, p = .294), we argue that in line with previous research on personalization (Li, 2016) and interactivity (Cauberghe, Geuens, & De Pelsmacker, 2011), it is more relevant to use consumers’ perceptions of these variables. To test our hypotheses, we analyzed the data using Hayes’ (2013) PROCESS macro (Model 3) with 1,000 bootstrap samples. We conducted three separate analyses for the three dependents. We used perceived tone of voice and perceived valence (both mean centered) as independent variables. Service type was dummy coded (cell phone provider (utilitarian) = 0, bar (hedonic) = 1). We also included all interactions.

Table 6.2. Unstandardized Regression Weights 6 Att PI WOM Perceived valence .443*** .365*** .201*** Perceived tone of voice .115** -.008 .136** Service type (1: bar) .390*** .701*** .403*** Perceived valence x tone of voice .125*** .124*** .072*** Perceived valence x service type .158*** .232*** .150** Perceived tone of voice x service type -.196* -.289** .110 Perceived tone of voice x perceived valence x service .001 .014 .092* type R² .590 .426 .314 Note: ***p≤.001; **p≤.010; *≤.050

There is a positive main effect of perceived message valence on Att (b=.443, p<.001), PI (b=.365, p<.001) and WOM (b=.201, p<.001). The results also show a significant interaction between message valence and tone of voice on Att (b=.125, p<.001), PI (b=.124, p<.001) and WOM (b=.072, p<.001)

(Table 6.2.). Simple slope analyses indicate that the slopes for Att (m-1SD= .298, m+1SD= .588), PI (m-

1SD= .221, m+1SD= .509) and WOM (m-1SD= .117, m+1SD= .285) for sWOM messages with a more factual tone of voice (observed at the mean +1SD) are steeper than those for the ones with a more emotional

155 Chapter 6: Don’t be so emotional! tone of voice (observed at the mean -1SD), indicating a stronger effect of message valence for more factual than for more emotional messages (see Figure 6.2.). H1a, b and c are supported.

Figure 6.2. Two-way interaction plot with word-of-mouth intention as dependent variable

Two-way interaction plot 5.500

5.000

4.500

4.000

3.500

3.000 WOM Intention 2.500

2.000 Negative Positive Factual 2.583 5.279 Emotional 2.981 4.348 Message valence

Next, we find a significant interaction between message valence and service type on Att (b= .158, p<

.001), PI (b= .232, p< .001) and WOM (b= .150, p= .001). The slope gradients indicate that for Att

(m-1SD= .600, m+1SD= .442), PI (m-1SD= .596, m+1SD= .364) and WOM (m-1SD= .350, m+1SD=

.200), the slopes for the more hedonic service is steeper than for the more utilitarian service, demonstrating a stronger effect of message valence for a more hedonic service than for a more utilitarian service. H2a, b and c are supported. Finally, we find a significant three-way interaction effect between message valence, tone of voice and service type on WOM (b= .092, p= .012), but not on Att

(b= .001-, p= .971) and PI (b= .014, p= .761). H3a and H3b are rejected. We further probed the results for WOM. For a more utilitarian service, the slope of message valence is significantly steeper for a more factual (observed at the mean +1 SD) compared to a more emotional message (observed at the mean

-1 SD) (m-1SD= .117, m+1SD= .283, p< .001), as expected in H3c. For a more hedonic service, the slope of message valence is also significantly steeper for a more factual message than for a more emotional message (m-1SD= .161, m+1SD= .541, p< .001). Unexpectedly, the slope is even steeper for the more hedonic than for the more utilitarian service. This finding indicates than for more hedonic services, it is even more important to use a more factual tone of voice than for a more utilitarian service.

156 Chapter 6: Don’t be so emotional!

H4a, b and c are thus rejected. Overall, H1, H2, and H3c are supported, while H3a, H3b and H4 are rejected.

Discussion and conclusion Previous research has demonstrated a positive effect of positive (compared to a negative) eWOM on consumer responses (e.g. Bae & Lee, 2011; Purnawirawan et al., 2015). The current research corroborates this on SNSs. In response to the call of Barger et al. (2016) to examine the effects of content and product factors on consumer engagement in social media, the current study focuses on the moderating impact of a factual vs emotional tone of voice and a utilitarian vs hedonic service type on this relationship in a social networking setting. The findings suggest that consumers evaluate sWOM differently depending on the tone of voice and the type of service evaluated. This implies that the effects of sWOM cannot simply be generalized across all messages and product types, which is in line with previous findings for online reviews (e.g. Willemsen et al., 2011). The findings of the current study add to the growing body of literature on the boundary conditions of message valence effects on responses by showing that sWOM message valence has more impact on attitude and PI toward a service and positive WOM intention when a more factual tone of voice (rather than an emotional one) is used.

Message valence also has a stronger effect on consumer responses when evaluating an sWOM message 6 about a more hedonic service compared to one about a moreutilitarian service. We conclude that these findings, previously documented in other fields (Park & Lee, 2009; Park & Lee, 2008; Wu & Wang, 2011), also hold in the context of SNSs.

Our findings also imply that matching tone of voice to the nature of a service does not lead to a greater impact of message valence on consumer responses than mismatching. In our study, for both more utilitarian and more hedonic services, the informative effect of a factual message seems to dominate the potential effect of a match in terms of the tone of voice. In other words, a more factual positive message leads to a more positive attitude toward the service provider, PI, and positive WOM intention, regardless of the product type. While this result was unexpected, it can be understood in light of other research. Klein and Melnyk (2016) argue that matching (or mismatching) a message for hedonic services might matter less due to the fact that arguments for this type of service might not be scrutinized in detail. They suggest that the presence of arguments offers a heuristic cue upon which decisions can be

157 Chapter 6: Don’t be so emotional! based. This seems especially true for WOM intention, as we found a significant three-way interaction here in the opposite direction of what we expected: more factual arguments seemed even more influential (compared to emotional ones) on WOM intention for a hedonic service than for a utilitarian service. Eisingerich et al. (2015) already posited that positive WOM intentions are more sensitive to social disapproval from others. A poor recommendation could harm the recommender’s credibility and reputation (Cheung et al., 2009) and consumers may be less willing to take a chance in offering positive

(s)WOM about a service as a poor decision in this respect does not only have repercussions for their own service experience, but may also reflect poorly upon them as a source of information (Dens, De

Pelsmacker, & Purnawirawan, 2015). This may be the reason why we find this effect on WOM intention, but not on attitudes or PI. The quality of hedonic services is harder to judge than that of utilitarian services. Review readers may only feel confident to recommend a (hedonic) service to others when there are clear positive factual arguments to do so.

Managerial implications

Our results provide guidelines for practitioners and influencers. First, message valence has a consistent effect on consumer responses: positive reviews lead to more positive responses. Service providers should thus encourage their customers to post as many positive reviews on SNSs as possible and avoid negative reviews to be posted. Second, messages that are formulated in a factual tone of voice reinforce the effect of review valence on the attitude toward the service provider, PI, and positive WOM intention.

Research suggests that consumers do not only spread eWOM out of personal interest (e.g., venting negative feelings or economic incentives), but also because they are concerned about others.

Consequently, we advise people posting sWOM and marketers soliciting positive sWOM to back up their sWOM with facts and avoid emotional argumentation. Third, our results imply that recipients of sWOM messages are generally more influenced by sWOM about hedonic services than utilitarian ones. It is, therefore, more beneficial for marketers of predominantly hedonic services to persuade potential customers via sWOM, and they should invest relatively more effort in trying to elicit positive sWOM or countering negative sWOM. Finally, there is no need for consumers or service providers to try to match their tone of voice of sWOM to the type of service. Regardless of the service type, factual messages have a greater impact than emotional ones. In terms of triggering WOM, the effect of factual review messages on consumers is even stronger for hedonic than for utilitarian services in terms of reinforcing

158 Chapter 6: Don’t be so emotional! the message valence effect. Consequently, in terms of customer engagement, this again underlines the importance of soliciting positive factual reviews for marketers of hedonic services, even more so than for providers of utilitarian services. Especially providers of hedonic services should, therefore, try to motivate satisfied customers to post more factual details about their positive experiences or they could ask for factual details when interacting with them. On their own site, they could provide a standard fill- out form or template to stimulate more factual consumer reviews, or they could more prominently highlight client testimonials that spontaneously describe factual aspects of their experience.

Limitations and suggestions for future research

The limitations of this study provide opportunities for future research. First, respondents were only exposed to a single sWOM message. Future research could examine the impact of a set of different sWOM messages about the same service. Kim and Gupta (2012) found that a single review with negative emotions decreases the negative impact of the review on product evaluations (compared to a non- emotional negative review), whereas multiple negative emotional reviews increase the negative impact (compared to multiple non-emotional negative reviews). Furthermore, for a positive emotional review, they found that one review did not impact product evaluations and multiple emotional reviews did. We also did not consider different types of emotions, nor did we use different types of attributes. Felbermayr 6 and Nanopoulos (2016) used data mining to examine the role of different emotions in online reviews for different product categories. Their findings already hint that there is an interaction between different emotions and product type. For example, for the category “tools and home improvement” joy is not a very important dimension to predict reviews helpfulness. For games, on the other hand, this is an important dimension. This might be a foundation upon which further experimental analyses might build. Furthermore, Teng et al. (2014) examine different aspects, such as strength, comprehensiveness, and relevance, in determining argument quality of online reviews, which future research could employ to distinguish between different types of arguments. Although we surveyed 400 respondents, which is in line with previous experiments in the field (e.g. Dens et al., 2015: 54 respondents per condition; Wang,

Cunnigham, & Eastin, 2015: 41 respondents per condition), future research could examine a larger sample. Some services appear to be hybrid with regard to buying motivations and are thus both hedonic and utilitarian. For example, our findings suggest that even though the bar was more utilitarian than hedonic, the hedonic score still scored above the midpoint of the scale. This might be the result of the

159 Chapter 6: Don’t be so emotional! fact that utilitarian services can also induce a person-specific experiences. Future research could examine services which is purely utilitarian or hedonic compared to more hybrid services to test the robustness of our findings.

Second, we did not examine the impact of the context. Our sWOM message was presented on a

Facebook wall with only one other message, which was intentionally kept neutral (profile picture update) and was kept constant over the different conditions. However, the mood induced by, e.g., the valence of other sWOM messages might also have an impact on the processing of sWOM messages. In line with the context appreciation theory, the context can be used as a source of information in the processing mechanism. It has been shown to positively impact likeability, informativeness and even brand recognition in the context of television advertising. A positive mood, as a result of the context, can be transferred to consumers’ responses (De Pelsmacker et al., 2002).

Third, Petty and Cacioppo (1984) indicate that attitudes formed via the central route are more time- resistant than those formed via the peripheral route and that people are more likely to act upon attitudes based on issue-relevant thinking. Future research could shed light on whether the use of a factual tone of voice and a hedonic buying motivation increases the likelihood of central processing in a longitudinal experiment.

Finally, our study only used self-reported attitudes and intentions and did not measure actual behavior.

The rise of social media has led to the development of data analytics, enabling researchers to analyze real social network data and examine the impact on, for example, sales (Rui et al., 2013). Measuring self-reports is common in the context of experiments, where people are often confronted with fictitious brands for reasons of internal validity (Geuens & De Pelsmacker, 2017). Adding behavioral data would strengthen the empirical validation of our results.

160

The Impact of Relational Characteristics on Consumer Responses to Word-of-Mouth on Social Networking Sites1, 2

1 Manuscript published as De Keyzer, F., Dens, N., & De Pelsmacker, P. (2019). The impact of relational characteristics on consumer responses to word of mouth on social networking sites. International Journal of Electronic Commerce, 23(2), 212-243. 2 An earlier version of this chapter was presented at 2015 Etmaal van de Communicatiewetenschap, Antwerp, Belgium. 7 Chapter 7: The impact of relational characteristics

Abstract

Previous research has consistently found an effect of the valence (positive vs. negative) of electronic word of mouth in general and of word of mouth on a social networking site (sWOM) specifically on consumer responses. The current study investigates how interpersonal and person-to-site relational characteristics (homophily, tie strength, and source credibility) moderate this effect on consumer responses to sWOM (behavioral and positive word-of-mouth intention). The results show that interpersonal homophily and source credibility both significantly reinforce the effect of sWOM valence on behavioral intention and positive word-of-mouth intention. Only considering person-to-site relational characteristics as antecedents, (person-to-site) homophily and source credibility reinforce the effect of sWOM valence on behavioral intention and on positive word-of-mouth intention. However, including both the interpersonal and the person-to-site relational characteristics as antecedents results in all person-to-site relational characteristics becoming nonsignificant as moderators. This study advances the sWOM literature by concurrently examining how both interpersonal and person-to site relational characteristics moderate the effect of message valence on sWOM responses. The findings imply that marketers should try to stimulate sWOM from credible sources that are homophilous to the target audience, as these relationships reinforce the positive impact of sWOM valence on behavioral intentions.

162 Chapter 7: The impact of relational characteristics

Introduction

It is difficult to imagine a world today without social networking sites (SNSs), applications such as

Facebook, Twitter, and LinkedIn “that enable users to connect by creating personal information profiles, inviting friends and colleagues to have access to those profiles, and sending e-mails and instant messages between each other” (Kaplan & Haenlein, 2010, p. 63). More than ever, SNSs play an important role in people’s lives, outgrowing their function as merely a medium or platform through which communication takes place to actually shaping conversations and becoming an actual “actor” in the communications process (Brown et al., 2007; Kim, Kandampully, et al., 2018). As a result of its significance in many people’s daily lives, Facebook, for example, has become one of the top 10 Best Global Brands (Interbrand, 2017).

Because of the popularity of SNSs, electronic word of mouth on social network sites (sWOM) has become increasingly important for both consumers and marketers (Chu & Kim, 2018). For example, 38 percent of online shoppers have posted messages on Facebook about products they used, and 62 percent of online shoppers read product-related messages from Facebook connections (Sociable Labs, 2012). sWOM influences sales, especially for services (Babić Rosario et al., 2016). One study found that 75 percent of people who read sWOM on Facebook actually click through to the retailer’s website (Sociable Labs, 2012). sWOM is expected to become increasingly influential since SNSs are easy to operate and do not require a lot of Internet-related knowledge (Hennig-Thurau et al., 2004). Despite their frequent 7 occurrence, consumers’ responses to sWOM are underresearched (Aghakhani, Karimi, & Salehan, 2018). sWOM about a product or service can be positive or negative. For example, a consumer can post a positive comment about a product or even recommend the product (“I really liked the comfy bed in my hotel room: if you’re looking for a place to stay, go there!”). On the other hand, a consumer can post a negative comment or even advise against buying it (“My hotel room was infested with cockroaches: don’t stay there!”). Previous research found that the valence of electronic word of mouth (eWOM) or sWOM strongly affects readers’ attitudes and behavioral intentions (Chu & Kim, 2018; Purnawirawan,

De Pelsmacker, & Dens, 2012). A positively (negatively) valenced message positively (negatively) impacts consumers’ attitudes (e.g. Lee, Park, & Han, 2008; Purnawirawan et al., 2015), purchase intentions (e.g. Bae & Lee, 2011), and even sales (Babić Rosario et al., 2016).

163 Chapter 7: The impact of relational characteristics

Although the effect of eWOM/sWOM valence appears to be quite consistent, moderating factors can reinforce or weaken this effect (e.g. Doh & Hwang, 2009; Kusumasondjaja, Shanka, & Marchegiani,

2012; Purnawirawan et al., 2015). WOM, both offline and online, does not take place in a social vacuum.

The relationship between the receiver of the message and the source of the message is part of the context in which the interaction takes place. Interpersonal relational characteristics are important to understand how people react to sWOM (Chu & Kim, 2011; Kim, Park, et al., 2018). SNSs are designed to build and maintain relationships (boyd & Ellison, 2007). In contrast to many review sites, people on

SNSs are identifiable. This also means that the sWOM reader can attribute an sWOM message to a specific sender and can, therefore, evaluate his or her interpersonal relationship with the sender, which may not be the case for other types of eWOM. Therefore, a better understanding of the effect of relational characteristics on the impact of sWOM is essential to both marketers and scholars (King,

Racherla, & Bush, 2014). Previous research has taken the first steps in uncovering the effects of interpersonal relational characteristics in the context of sWOM (Aghakhani et al., 2018; Brown et al.,

2007; Chu & Kim, 2011). For example, Chu and Kim (2011) investigated the effects of interpersonal homophily, tie strength, trust, normative and informative influence on (general) opinion seeking, opinion giving, and pass-along behavior in SNSs.

Important to note, based on qualitative research, Brown et al. (2007) argued that, in online social networks, individuals develop relationships with the (social network) site on which they are communicating rather than with other individuals using the site. As a result, they claimed that sWOM responses are influenced not so much by the interpersonal relational characteristics between the sender and the receiver but rather by the relational characteristics between the SNS and the receiver of the message (person-to-site relational characteristics). This proposition could represent a major step forward in theory building on sWOM effects, as prior research has almost exclusively focused on message and interpersonal aspects, largely ignoring effects of the platform. At the same time, as mentioned, the goal of SNSs is to build relationships with and between people. The proposition of Brown et al. (2007) that the site relationship trumps interpersonal factors, therefore, comes across as somewhat counterintuitive. Empirical validation of the proposition would aid to contribute to the debate.

As a result, there is clearly a need to empirically test the proposition of Brown et al. (2007). Recently,

Kim, Kandampully, et al. (2018) offered a partial test of this proposition by studying the effects of

164 Chapter 7: The impact of relational characteristics person-to-site homophily, tie strength, and source credibility on “sWOM effectiveness” (the ability of sWOM to influence purchase decisions). They showed that the person-to-site variables indeed explain some of the variance in sWOM effectiveness. At the same time, they use only person-to-site variables without taking interpersonal relational characteristics (or other relevant variables) into account, which makes it an invalid, or at least incomplete, test of Brown et al.’s (2007) proposition.

Based on the framework provided by, this study focuses Brown et al. (2007) on three relational characteristics: homophily (the extent to which a sender and receiver are alike), tie strength (the strength of relations), and source credibility (the trustworthiness and expertise of the source) (Brown &

Reingen, 1987; Money, Gilly, & Graham, 1998). We test the extent to which these three characteristics, defined both interpersonally and with respect to person-to-site, moderate the influence of sWOM valence on recipients’ behavioral intention (in this case, intention to watch a movie) and positive word- of-mouth (PWOM) intention (intention to positively mention the product [movie] to others).

The contribution of the current study is thus that it explicitly tests and challenges the proposition of Brown et al. (2007) and its empirical validation by Kim, Kandampully, et al. (2018). Furthermore, we contribute to theory development by suggesting theoretical frameworks to explain why person-to-site relational characteristics would influence sWOM responses. Building on these two studies, our study allows for an empirical assessment of the relative importance of both interpersonal and person-to-site relational characteristics (homophily, tie strength, and source credibility) for responses to sWOM 7 messages. To our knowledge, our study is the first to do this. Moreover, we test both a positively and a negatively valenced message and analyze how the relational characteristics under study moderate the effects of sWOM valence. Many prior studies on eWOM/sWOM have focused on either purely positive or purely negative WOM (Lee et al., 2008) and considered the relational characteristics as direct antecedents to sWOM responses. Advertising practitioners may also use these insights to stimulate sWOM from certain types of people or on certain types of sites in order to use social media more effectively.

165 Chapter 7: The impact of relational characteristics

Literature Review and Hypotheses

The Influence of Message Valence on Consumer Responses to sWOM

Social media are used to communicate both positive and negative sWOM. As mentioned earlier, the valence effect has been consistently found in previous research. For example, with regard to social media research, Rui et al. (2013) found that positive tweets about a movie positively influence movie sales, whereas negative tweets have a negative effect. At the same time, research documents that, for example, consumer characteristics such as brand familiarity (Purnawirawan et al., 2015) or sender identification (Kusumasondjaja et al., 2012) can play a moderating role. One potential category of moderators that is not extensively explored yet in the context of sWOM is the relation between an sWOM sender and its receiver.

As mentioned, the purpose of this paper is to study the moderating effects of homophily, tie strength, and source credibility on the relationship between sWOM valence and consumer responses. Unlike previous papers, we hereby consider these relational characteristics from both an interpersonal and a person-to-site perspective. We begin by developing hypotheses for the interpersonal relational characteristics.

The Role of Interpersonal Relational Characteristics

Homophily

Interpersonal homophily is the extent to which pairs of individuals are alike and share, for example, the same age group, gender, education, lifestyle, social class, or interests (Brown & Reingen, 1987; Rogers

& Bhowmik, 1970). Mcpherson, Smith-lovin, and Cook (2001) posited that similarity in sex, age, religion, education, race, and ethnicity structure our interpersonal relationships. Even though some cues that are present in an offline context are not always available in online relationships, research has found that, for instance, perceptions of interests and opinions are used to evaluate interpersonal homophily online

(Blanton, 2001; Gilly et al., 1998).

Research in an offline context has indicated that perceived interpersonal homophily increases the likelihood of perceiving the other as being more persuasive (Gilly et al., 1998). According to Lazarsfeld and Merton (1954), this is because people who are similar are more likely to interact with each other in comparison to people who are dissimilar. Therefore, information is mostly exchanged between

166 Chapter 7: The impact of relational characteristics individuals who are homophilous (Rogers, 1995; Rogers & Bhowmik, 1970). Festinger’s (1954) social comparison theory posits that people implicitly assume that individuals similar to themselves have similar needs and preferences, and thus they tend to compare themselves with others who are similar. Therefore, perceived interpersonal homophily stimulates a greater level of interpersonal attraction, trust, and understanding (Rogers & Bhowmik, 1970; Ruef, Aldrich, & Carter, 2003), as it can serve as a cue to indicate that the product or service is suited for people “like them” (De Bruyn & Lilien, 2008). Moreover, perceived interpersonal homophily stimulates the diffusion of information about products and services (Yang, Tang, Dai, Yang, & Jiang, 2013). For these reasons, perceived interpersonal homophily has a positive influence on the chance that information is used. As a result, information from a homophilous source has more influence in the decision-making process compared to information from a heterophilous (dissimilar) source (Gilly et al., 1998; McCroskey, Hamilton, & Weiner, 1974; Pentina, Ainsworth, & Zhang, 2018; Steffes & Burgee, 2009).

Research on the effect of interpersonal homophily in an online context has extended these findings.

Wang, Walther, Pingree, and Hawkins (2008) showed that the stronger the perceived interpersonal homophily in Websites and online discussions groups, the more likely people are to adopt the advice provided there. Steffes and Burgee (2009) used the eWOM forum RateMyProfessors.com to examine the impact of social ties on eWOM. Not only were students more likely to engage with homophilous sources, the information they provided was also more likely to influence their decision making than 7 heterophilous sources. In a study on the review site Yelp, Pentina et al. (2018) found that perceived interpersonal homophily positively influences perceived perceptions of the review message (e.g., helpfulness). Based on this evidence, we expect that interpersonal homophily will reinforce the effects of sWOM valence. In general, positive sWOM should result in more positive consumer responses than negative sWOM. If homophilous sources are indeed more influential than heterophilous sources, responses to positive sWOM should be more positive and responses to negative sWOM more negative when the sWOM is posted by more homophilous sources. In other words, the differential effects of positive and negative sWOM should become greater with increasing interpersonal homophily. Therefore, we expect the following:

167 Chapter 7: The impact of relational characteristics

H1: Interpersonal homophily moderates the effect of sWOM valence on receivers’ (1) behavioral intention and (2) PWOM intention: The more the sender is perceived by the receiver as homophilous, the stronger the effect of sWOM valence.

Tie Strength

Interpersonal tie strength is “a multidimensional construct that represents the strength of the dyadic interpersonal relationship in the context of social networks” (Money et al., 1998, p. 79). It is characterized by the importance attached to the social relation, the frequency of social contact, the type of social relation, and the intimacy between two parties (Brown et al., 2007; Granovetter, 1973; Marsden

& Campbell, 1984). In a social networking context, interpersonal tie strength could be reflected, for instance, by the number of common friends (Fogués, Such, Espinosa, & Garcia-Fornes, 2014; Tuna et al., 2016), shared activities (Chu & Kim, 2011), recency of communication (Gilbert & Karahalios, 2009), or the number of interactions between two people (Chu & Kim, 2011; Gilbert & Karahalios, 2009).

Previous research has found that a strong tie between a sender and a receiver has more impact on the receiver’s behavior than a weak tie (e.g. Brown & Reingen, 1987; Chu & Kim, 2011). De Bruyn and

Lilien (2008) showed that tie strength has an impact on opening a received e-mail: When the tie is stronger, people are more likely to open the e-mail. Furthermore, the strength of a tie affects the information flow: Individuals in a strong tie relationship interact more frequently and exchange and spread more information (Brown & Reingen, 1987). Moreover, information from a strong tie is perceived as more trustworthy, and can therefore reduce potential risks (De Bruyn & Lilien, 2008; Rogers, 1995).

As a result, strong ties are more influential than weak ties. In a study on Facebook, Bitter and Grabner-

Kräuter (2016) found that readers’ visiting intentions in response to positive sWOM were significantly higher when the message was sent by a strong tie. They suggested that tie strength acts as a reference point for Facebook users in their evaluation of brand-related information (Bitter & Grabner-Kräuter,

2016). The sWOM from a strong tie will result in more in-depth processing of the information in the message (Bitter & Grabner-Kräuter, 2016). Therefore, we posit that perceived interpersonal tie strength should reinforce the effect of sWOM valence in such a way that sWOM coming from a strong tie will have a stronger impact on consumer evaluations than sWOM coming from a weak tie.

168 Chapter 7: The impact of relational characteristics

H2: Interpersonal tie strength moderates the effect of sWOM valence on receivers’ (1) behavioral intention and (2) PWOM intention: The more the sender is perceived by the receiver as a strong tie, the stronger the effect of sWOM valence.

Source Credibility Source credibility refers to the perception of “a message sender’s positive characteristics that influence receivers’ acceptance of the message communicated” (Uribe, Buzeta, & Velásquez, 2016, p. 4406). Source credibility consists of two dimensions: expertise and trustworthiness (Ohanian, 1990). Sources that are perceived as competent in a certain matter (e.g., based on knowledge on or experience with a certain topic, occupation, or social training) and/or perceived as trustworthy (e.g., not acting out of self- interest) will be considered as more credible (Ohanian, 1990). Source credibility, in turn, causes receivers to pay more attention to the message, perceiving it as more useful (Li et al., 2013), reliable

(Cheung et al., 2008), and credible (Cheung et al., 2009), making the message more persuasive (Teng et al., 2014). For example, people often seek advice from experts because the experts are expected to possess more knowledge about a certain topic and be able to provide accurate information

(Pornpitakpan, 2004). As a result, these experts are more likely to influence consumers’ decisions than nonexpert sources(Herr et al., 1991). Therefore, we expect that interpersonal source credibility moderates the effect of sWOM valence: An sWOM originating from a source that is perceived as more credible should have a stronger impact on consumer responses than an sWOM originating from a source 7 that is perceived as less credible:

H3: Interpersonal source credibility moderates the effect of sWOM valence on receivers’ (1) behavioral intention and (2) PWOM intention: The more the sender is perceived by the receiver as credible, the stronger the effect of sWOM valence.

The Role of Person-to-Site Relational Characteristics

People’s Relationships with Brands

Brown et al. (2007) suggested that people use the Website (rather than people or their avatars) as a social proxy for individual identification. This implies that the “source” for online information is not the individual that posts it but the Website on which it is posted. This has recently been confirmed by Kim,

Kandampully, et al. (2018). They concluded that “even without face-to- face interactions, consumers

169 Chapter 7: The impact of relational characteristics still develop various types of relationships with websites, the strength of which has a strong influence on their evaluations of the website, its reviews and ultimately, their purchase decisions” (Kim,

Kandampully, et al., 2018, p. 251). Brown et al. (2007) and Kim, Kandampully, et al. (2018) drew on theories in social psychology, branding, and media equation theory to underpin this proposition. Social psychologists and advertisers have long acknowledged that inanimate objects (such as brands) can be associated with human characteristics. The reasoning behind this is that these objects can be associated with personality traits that help consumers express themselves or that are symbolic for the consumer

(Aaker, 1997). In the marketing literature, it is widely accepted that consumers develop relationships not only with other people but also with brands, websites, or SNSs (Aaker, 1997; Fournier, 1998). For instance, brand personality, a set of human characteristics associated with a brand (Aaker, 1997), is used to describe a brand as a partner in a relationship (Sweeney & Brandon, 2006). Self-congruity theory posits that consumers prefer brands (and, by extension, sites) that are congruent to their own personality (Aaker, 1997). Congruity between an individual’s characteristics and those of a brand, store, or product could therefore improve attitudes, preferences, and even behavior (Helgeson & Supphellen,

2004). People thus also compare themselves to brands and other objects (e.g.,, holiday destinations) to develop perceptions of how well the object fits with their own self-concept, a process referred to as self-congruity . For example, Koo, Cho, and Kim (2014) found that self-congruity with an online store positively affects the evaluation of visual and information atmospheric cues of the online store, which causes delight and consequently increases purchase intention. Brown et al. (2007) found evidence for the humanization of websites (see also the upcoming section), and this has been confirmed by Kim,

Kandampully, et al. (2018). Website personalities share attributes with human personality due to their interactive features (dialogue and customization), as well as with brand personality due to the association with the brand or organization that owns the Website (Chen & Rodgers, 2006). Finally, media equation theory suggests that interactions with media (e.g., SNSs) are similar to interactions in real life (Brown et al., 2007). SNSs can therefore also be seen as actors in the evaluation of sWOM.

Homophily

Brown et al. (2007) defined person-to-site homophily as “the congruence between a user’s psychological attributes and the website content, such as shared group interests and a shared mindset between a user and the site” (p. 10). Here, content refers to the actual textual content (i.e., the information

170 Chapter 7: The impact of relational characteristics content) of the site rather than who actually provides that information (i.e., the individual users).

According to Brown et al. (2007), shared group interests reflect a match between the information seeker’s own interests and the content provided by the SNS. For example, a consumer who is very much interested in movies might find a movie review Website to represent “shared interests.” A broad range of information and lack of specificity tend to contribute to a greater feeling of shared interests. More specifically, sites with a broad range of relevant content are seen as more likely to introduce unexpected things of interest. Also, Brown et al. (2007) stated that SNSs that are more general tend to attract users because they know that they will be able to find an issue that engages them. Shared group mindset is based on psychological similarity, members perceiving the online social network as a unit that thinks and feels in a convergent way. In the study by Brown et al. (2007), this is illustrated through gratifications (posts thanking other contributors, showing appreciation, and generally supporting the consensus opinion) and collective postings (using collective words such as “our,” “we,” and “us”).

The results from the qualitative research by Brown et al. (2007) indicate that, in online social networks, respondents’ social affiliations that display homophily are with the Website rather than with individuals.

Kim, Kandampully, et al. (2018) found that when a review site provides information consistent with a consumers’ own interests the Website is rated as more homophilous. Moreover, consumers tend to be more positive about a homophilous review site and reviews presented on the site. Drawing on literature on the relationship between individuals and brands, Websites, and media, and based on Brown et al.’s 7 (2007) proposition and Kim et al.’s (2018) findings, it could be expected that a similar moderating effect of person-to-site homophily occurs as that of interpersonal homophily in H1:

H4: Person-to-site homophily moderates the effect of sWOM valence on receivers’ (1) behavioral intention and (2) PWOM intention: The more the SNS is perceived by the receiver as homophilous, the stronger the effect of sWOM valence.

Tie Strength

Brown et al. (2007) proposed that the idea of individual-to-individual social ties is less relevant in an online environment than offline. In their qualitative study, none of their respondents explicitly mentioned any type of “interpersonal relationship” between themselves as information seeker and another individual as the information source. Rather, they suggested that the online information “source” is the

171 Chapter 7: The impact of relational characteristics site. They described person-to- site tie strength as “the intensity of an interactive and personalized relationship between an individual and a website” (p. 10). For example, they documented that people refer to Websites in ways as “… like they know me” or “always understand me.” Brown et al. (2007) characterized person-to- site tie strength by dimensions such as Website reciprocity (“interaction”) and emotional closeness. Interaction is influenced by regular e-mails and updates automatically generated by the Website. Kim, Kandampully, et al. (2018), following Brown et al. (2007), concluded that individuals might form strong ties with a site that they frequently visit and perceive as important. Brown et al. (2007) concluded that people want to develop close relationships with online community Websites

(rather than with individuals in the community). For example, one respondent commented, “I like the recommendations ’cos it makes me feel like they know me.” Their study also provides evidence of a

Website relationship through a substantial number of collective rather than individual posts together with the “humanization” attributed to the Website (e.g., “You always understand me”).

Kim, Kandampully, et al. (2018) found that consumers with a strong relationship with a review site indicated more favorable attitudes toward the site itself as well as the content provided on it. Pentina,

Gammoh, Zhang, and Mallin (2013) found that people who perceived their tie with the SNS as strong were more likely to visit websites of brands hosted on the SNS, to purchase from these sites, and to recommend the brands to their friends and acquaintances. Building on the balance theory, they advocated that the perceived tie strength with the SNS can be transferred to brands active on the SNS and, in turn, can influence behavioral intentions toward these brands (Pentina et al., 2013). As a result, we expect that perceptions of tie strength with the SNS can transfer to the perceptions of a brand mentioned in an sWOM message and, thus, can influence behavioral intentions toward that brand, such as a movie that is recommended or advised against in sWOM:

H5: Person-to-site tie strength moderates the effect of sWOM valence on receivers’ (1) behavioral intention and (2) PWOM intention: the more the SNS is perceived by the receiver as a strong tie, the stronger the effect of sWOM valence.

Source Credibility

Brown et al. (2007) and Kim, Kandampully, et al. (2018) mainly ascribed online source credibility to

Website factors rather than people. They stated that the credibility of a Website as a source of

172 Chapter 7: The impact of relational characteristics information reflects “the perceived competence of the site and its membership, characterized by the site’s trustworthiness and its actors’ expertise” (Brown et al., 2007, p. 10). Trustworthiness is determined by the site’s perceived intentions (Kim & Kang, 2014): For example, is it an “independent” site seeking to inform, or a brand-sponsored site seeking to persuade? Prior experience with the site also influences how people evaluate the credibility of information from a site. Kim, Kandampully, et al. (2018) confirmed that consumers evaluate the credibility of a review based on perceptions of credibility of the review site. A common theme in the results of Brown et al. (2007) was concerned with some kind of “authority” that the Website could generate, which then gave any information on that site more weight. If a Website is not perceived as credible, the information and services will not be trusted and information needs will be filled elsewhere (Fogg, 2003). The effect of sWOM valence will, in that case, be weaker:

H6: Person-to-site source credibility moderates the effect of sWOM valence on receivers’ (1) behavioral intention and (2) PWOM intention: The more the SNS is perceived by the receiver as credible, the stronger the effect of sWOM valence.

The Relative Importance of Interpersonal versus Person-To-Site Relational Characteristics

Next, to formally testing the hypotheses, the analysis will also allow us to investigate the key question of the current study, that is, the relative importance of interpersonal and person-to-site relational 7 characteristics as moderators of the effect of sWOM valence on movie watching and PWOM intention.

As mentioned, J. Brown, et al. [10] claimed to have “strong evidence” that, in online social networks, people behave as if the platforms themselves are primary “actors” and that online communities (not individual users) act as a social proxy for individual identification. Moreover, they argued that it is the relations to the Website that should be considered as the basis for sWOM effectiveness, as relations with individual actors on the site are “not particularly relevant.” For example, they wrote,

Our findings suggest that homophily of an interpersonal relationship, as based on an evaluation of individual characteristics, is not particularly relevant in an online context. Rather, the findings suggest that it is notions of shared group interests and group mind-set, evaluated at the level of the Web site itself, which drive online homophily. (p. 9)

173 Chapter 7: The impact of relational characteristics

Similar statements are made for tie strength and source credibility. In the context of the current study, that would imply that the extent to which the person-to-site variables moderate the effects of sWOM valence (H4–H6) should be greater than the extent to which interpersonal relations (H1– H3) do.

We argue that J. Brown et al. [10] overvalue the importance of person-to-site relation characteristics relative to interpersonal relationships. Other than their own study, there is little evidence to support the idea that individuals develop person-to-site relations rather than interpersonal ones. While Kim,

Kandampully, et al. (2018) supported the idea that person-to-site relational characteristics influence sWOM effectiveness, theirs is the only empirical study, and they do not test the relative importance of these person-to-site relational characteristics compared to interpersonal characteristics. On SNSs, people are identifiable, and individuals therefore have plenty of access to information to evaluate homophily, tie strength, and source credibility of the sender. There is plenty of evidence that interpersonal relational characteristics determine the effectiveness of eWOM and sWOM, as described in the development of H1 to H3. If the proposition of J. Brown et al. [10] were true, none of these studies should have found any effect of interpersonal variables. Given the nature of SNSs, namely, to build relationships with and between people, we believe that individuals will be influenced by which person sends out the sWOM rather than what platform it is on. After all, the information content is created by an identifiable, individual user, not the platform. Although we do not contest that people do also form relations with the platform, and this, too, can influence sWOM responses (as developed in

H4–H6), we argue that those relations are subordinate to the interpersonal relations. When it comes to sWOM, people will still often consider the platform as just the medium, and rather see the individual sender as the “sender.” As such, the influence of the site should be smaller than that of the sender. We expect the following:

H7: The moderating influence of the person-to-site relational characteristics on the effect of sWOM valence on a receivers’ (1) behavioral intention and (2) PWOM intention is weaker than that of interpersonal relational characteristics.

The conceptual framework and hypotheses are summarized in Figure 7.1.

174 Chapter 7: The impact of relational characteristics

Empirical Study

Design

To test our conceptual framework (Figure 7.1), we created a 2 (sWOM valence: positive vs. negative)

× 2 (interpersonal homophily: low vs. high) × 2 (interpersonal tie strength: low vs. high) × 2 (sender source credibility: low vs. high) full-factorial between subjects experiment. Due to the popularity of

Facebook as an SNS, we set our study on Facebook. Worldwide, each day on average 1.45 billion users visit Facebook (Facebook, 2018), and it accounts for 42 percent of all social media visits in the United

States (Statista Inc, 2016).

We designed 16 versions of a fictitious (noninteractive) Facebook profile (see Appendix A.F) including an sWOM message. Because the Facebook profile was drafted for a fictitious person (the sWOM sender), we manipulated interpersonal homophily (low vs. high), interpersonal tie strength (low vs. high), and sender source credibility (low vs. high) to induce variance in perceived interpersonal homophily, tie strength, and source credibility. Person-to-site homophily, tie strength, and source credibility were not manipulated but

Figure 7.1. Conceptual Framework

Interpersonal Interpersonal Interpersonal Homophily Tie Strength Source Credibility 7

H1 H2 H3

a. movie watch sWOM intention valence b. positive WOM intention

H4 H5 H6

Person-to-site Person-to-site Person-to-site Source Homophily Tie Strength Credibility

175 Chapter 7: The impact of relational characteristics measured (described later), because Facebook is a site with which most participants are highly familiar and for which people naturally vary in their perceptions of the three variables under study. The sWOM valence was manipulated by a short post on the profile wall recommending (positive: Go see it, it’s very good!) or advising against (negative: Don’t go see it, it’s not very good.) the movie Dirty. Due to the fact that six potential moderating variables are included in the model we examine the perceived measures to avoid potential differences in the manipulation of the actual variables. The experiment took place before this American thriller was officially released in the United States to rule out that participants had already seen the movie prior to the experiment. Eight respondents who reported to have seen the movie were filtered out of the data set.

To manipulate interpersonal homophily, the profile pages were adapted on four aspects (age, gender, ethnicity, and interests) based on research by Thelwall (2009) and Mcpherson et al. (2001). In the highly homophilous condition, these four aspects were all consistent with what respondents indicated applied to themselves earlier in the questionnaire. For example, a woman 22 years old of Hispano-

American ethnicity interested in basketball and dancing would be exposed to a profile page of a

(fictitious) Hispano-American woman 25 years old (held constant for all “high homophily” conditions, as all respondents were between 22 and 36 years old) that included a dancing and basketball page as two of her “page likes.” In the low-homophily condition, we made sure that the profile of the sWOM sender was substantially different. The profile was kept constant across the low-homophily conditions, with an age of 55, of Chinese ethnicity (Asian Americans were therefore excluded from the sample), with more

“general” page likes (Facebook and Coca-Cola). The profile was also of the opposite sex as the respondent’s.

To manipulate interpersonal tie strength, we differentiated three aspects based on Marsden and

Campbell (1984), Liu-Thompkins (2012) and Gilbert and Karahalios (2009): interaction frequency, the number of common friends, and interaction recency. First, in the instructions immediately preceding exposure to the Facebook profile, participants were told that the profile was either that of an acquaintance with whom the respondent rarely interacted (low tie strength condition) or that of a close friend with whom they interacted almost daily (high tie strength condition). Second, our research used one common friend to manipulate the low tie strength condition and 35 common friends to manipulate

176 Chapter 7: The impact of relational characteristics the high tie strength condition. Finally, the profile page included an open Facebook chat box that read either “Hi! It’s been a long time! Want to grab a drink this Friday to catch up?” (low tie strength condition) or “Hi! Do you want to grab a drink this Friday? Same place, same time as last week?” (high tie strength condition).

Finally, we used four dimensions to manipulate sender source credibility, based on Purnawirawan, Dens, and De Pelsmacker (2014). In the high source credibility condition, the sWOM sender had studied film studies, liked movie-related pages (HBO, Netflix, and Inception), was a member of a film club, and mentioned that the focal movie (Dirty) was the 50th he or she had reviewed. In the low credibility condition, the sWOM sender had studied law; liked more general, non-movie-related pages (CBS News,

Stephen King, and Walmart); was a member of a book club; and mentioned that the focal movie was the first he or she had reviewed.

Data Collection Respondents (n = 801) were actual American Facebook users between the ages of 22 and 36 recruited for an online survey via a professional recruitment service. Participants could proceed through the questionnaire at their own pace. The questionnaire started with a welcome screen informing respondents that their answers would be processed in full anonymity and used in the context of academic research. Participants indicated whether they had a Facebook account and those who did not were redirected to the end of the questionnaire. The remaining participants were asked to complete 7 sociodemographic information (gender, birth year, education, ethnicity, and interests), which were used to describe the sample and (with the exception of education) manipulate interpersonal homophily. Next, as an introduction, they were asked how many days and how many minutes per day they had used

Facebook in the past week, together with a scale to measure Facebook intensity (three items, α = .840;

De Keyzer, Dens, & De Pelsmacker, 2017). Next, respondents rated their degree of perceived homophily with Facebook (four items, α = .901; McCroskey, McCroskey, and Richmond (2006) and Kim,

Kandampully, et al. (2018), perceived tie strength with Facebook (three items, α = .892; Kim,

Kandampully, et al., 2018; Shan & King, 2015), and source credibility of Facebook (five items, α = .870;

Ohanian, 1990, p.; see Table 1 for construct items and reliabilities).

177 Chapter 7: The impact of relational characteristics

The average age of the respondents was 27.97 (SD = 3.62). Of the sample, 53.1 percent were male and 43.7 percent were at least undergraduates. On the 7-point Facebook Intensity scale (described later), respondents scored 4.21 on average (SD = 1.71). Moreover, 56.6 percent of respondents indicated that they used Facebook every day. A majority (52.7 percent) used Facebook less than 30 minutes per day. Respondents were then randomly exposed to one of the 16 fictitious (noninteractive)

Facebook profiles with the instruction to imagine that this was an actual profile of one of their Facebook contacts. They were asked to look at the profile attentively and to try to get a picture of what this person would be like: what kind of things he or she was interested in, what kind of people he or she interacted with, and so on. We then measured the perceived valence of the sWOM message (one item;

De Keyzer et al., 2017). Next, participants rated their perceived interpersonal homophily with (three items, α = .899; McCroskey et al., 2006), perceived interpersonal tie strength with (three items, α =

.902; Shan & King, 2015), and perceived source credibility of (five items, α = .860; Ohanian, 1990) the

(fictitious) sWOM sender. Finally, we measured their behavioral intention (intention to watch the movie

[MWI]; one item; De Keyzer et al., 2017) and PWOM (three items, α = .807; De Keyzer et al., 2017).

All constructs were measured by means of 7-point Likert scales or semantic differentials. The means and standard deviations of these measures are shown in Table 7.1.

Table 7.1 shows that all Cronbach alphas are above .807. Per construct, mean scores across items were used in further analyses. As can be seen in Table 7.1, there is substantial variation in both the person- to-site and interpersonal measurements. Furthermore, the variance induced by the interpersonal manipulations is similar to the variance naturally occurring in the person-to-site aspects. Prior research indicates that actual and perceived attributes do not automatically match (Rogers & Bhowmik, 1970).

As perceptions are more likely to drive responses than actual attributes (McCroskey et al., 1974; Rogers

& Bhowmik, 1970), we used the measured scores on perceived interpersonal and person-to-site homophily, tie strength, and source credibility in the analyses instead of the manipulated conditions.

Confirmatory Factor Analysis

We first performed confirmatory factor analysis on the six interpersonal and person-to-site variables, using maximum likelihood estimation in AMOS 22. Indices of model fit indicate an acceptable fit of the confirmatory factor analysis (CFA) model (χ2/df= 4.639), comparative fit index and Tucker–Lewis index

178 Chapter 7: The impact of relational characteristics were over .90 (i.e., comparative fit index = .926, and Tucker–Lewis index = .917), root mean square error of approximation (.067) was below .07 (Hair et al., 2010) and adjusted goodness of fit index (.840) was over .80 (Hu & Bentler, 1999). Table 7.1. shows that the factor loadings for all indicators are large and significant, providing strong evidence of convergent validity. Also, the average variance extracted

(AVE) for each factor was above .50. Composite reliability estimates range between .833 and .928, which are well over the recommended .70 (Hair et al., 2010). To ensure discriminant validity, we tested if the square root of the AVE of a construct is greater than the correlation of that construct with all other constructs (Fornell & Larcker, 1981). Table 7.2. shows that this is the case: The diagonal shows the square root of the AVE per construct, and the off-diagonals show the correlations between each pair of constructs. No correlation was found to be higher than the square root of the AVE. Moreover, Table 7.2. shows that the average

179

180 Table 7.1. Measures C

hapter

Construct Items Mean Factor Cronbach’s Composite AVE

(SD) Loadings Alpha Reliability

7: The impact of 7: impact The Perceived interpersonal This person 3.89 .899 .906 .764 homophily (HM)  shares my values. (1.18) .765  is like me. .933  has a lot in common with me. .914 Perceived interpersonal tie  I want my relationship with this person to last for a long time. 3.38 .840 .902 .904 .759 strength (TS)  I feel strongly linked to this person. (1.31) .869

 The relationship with this person is important to me. .904

relational characteristics relational Perceived source credibility of This person is 4.70 .916 .928 .650 person (SC)  unreliable - reliable. (1.05) .681  untrustworthy - trustworthy. .632  not an expert - expert. .832  inexperienced - experienced. .842  unknowledgeable - knowledgeable. .880  unqualified – qualified .895  unskilled - skilled .842 Perceived person-to-site  Facebook shares my values. 3.48 .848 .901 .912 .727

homophily (FBHM)  Facebook is like me. (1.32) .956  Facebook has a lot in common with me. .950  My interests are similar to the content I can find on Facebook. .611 Perceived person-to-site tie  I feel strongly linked to Facebook. 3.52 .929 .892 .898 .747 strength (FBTS)  The relationship with Facebook is important to me. (1.55) .903  There are a lot of activities/interaction between Facebook and me. .751 Perceived person-to-site source Facebook is 4.29 .918 .918 .618 credibility (FBSC)  Unreliable – reliable (1.14) .708  Untrustworthy – trustworthy .602  Not an expert – expert .768  Inexperienced – experienced .799  Unknowledgeable – knowledgeable .872  Unqualified – qualified .881  Unskilled - skilled .836 Perceived sWOM Valence  This message is negative – positive. 4.09 - - - - (1.95) Movie Watch Intention (MWI)  If I were to choose a movie, I would consider Dirty. 3.36 - - - - (1.46) Positive Word-Of-Mouth  I am likely to say negative – positive things about Dirty to other 3.78 .821 .807 .833 .630 Intention (PWOM) people. (1.02)  I am not likely – likely to recommend Dirty to a friend or colleague. .622

 I am likely to discourage – encourage friends and relatives to go see Dirty. .910 Facebook Intensity  Facebook is part of my everyday activity. 4.21 - .840 - -  I feel out of touch when I haven’t logged onto Facebook for a while. (1.71)  I would feel sorry if Facebook shut down. Days spent on FB  In the past week, on average, approximately how many days have - - - - - you used Facebook? Minutes per day on FB  In the past week, on average, how many minutes per day have you - - - - - spent on Facebook?

Chapter 7: The impact of relational chara of relational 7: impact The Chapter

teris

181

ics

Chapter 7: The impact of relational characteristics shared variance is larger than the maximum shared variance. Thus, we can conclude that discriminant validity is confirmed (Hair et al., 2010). Nevertheless, examining the bivariate correlations between the constructs, we find that the correlations between homophily and tie strength for both interpersonal and person-to-site relational characteristics are above .60, which is rather high. This might result in multicollinearity issues that need to be taken into account in the analysis (see the following section).

Table 7.2. Square root of average variance extracted and correlations per factor

AVE MSV HM TS SC FBHM FBTS FBSC PWOM Perceived interpersonal homophily (HM) .764 .424 .874 Perceived interpersonal tie strength (TS) .759 .424 .651 .871 Perceived source credibility of person (SC) .650 .156 .392 .395 .806 Perceived person-to-site homophily .727 .599 .349 .305 .274 .853 (FBHM) Perceived person-to-site tie strength .747 .599 .310 .266 .222 .774 .865 (FBTS) Perceived person-to-site source credibility .618 .249 .276 .199 .359 .499 .481 .786 (FBSC) Positive Word-Of-Mouth Intention .630 .042 .123 .200 .176 .206 .093 .166 .794 (PWOM) Note: the square root of AVE can be found on the diagonal, the correlations are in the off-diagonals.

Common Method Bias Analysis

To test for common method bias, we first performed Harman’s single factor test. The first factor of our explanatory factor analysis accounts for only 33.237 percent of the variance explained, indicating that no common method bias is present (Podsakoff, MacKenzie, Lee, & Podsakoff, 2003). We further tested for common method bias by adding an unmeasured latent factor to the original CFA model. Then the standardized regression weights from this model were compared to the standardized regression weights of the original model without the common latent factor (Podsakoff et al., 2003). Only one difference over .200 was found. Due to the rather small changes in the standardized regression weights, we can say that adding the unmeasured latent factor did not change the original CFA model and, thus, we can assume that no common method bias was present (Podsakoff et al., 2003).

Manipulation Checks

Respondents rated the positive sWOM (M = 5.44, SD = 1.08) as significantly more positive than the negative sWOM (M = 2.79, SD = 1.69), t (689.224) = −26.513, p < 0.001. We can therefore conclude that our manipulation of sWOM valence was successful. The score for perceived homophily was significantly different between the high-homophily (M =4.01, SD =1.23) and the low-homophily (M

=3.76, SD =1.13), t(799) = −2.995, p =0.003, conditions. Furthermore, the score for perceived tie

182 Chapter 7: The impact of relational characteristics strength was significantly different between the high tie strength (M =3.52, SD = 1.35) and the low tie strength (M =3.24, SD = 1.25), t(799) = −2.989, p = 0.003, conditions. Finally, respondents rated the highly credible source conditions (M =5.05, SD = 1.045) as significantly more credible than the lowly credible source conditions (M =4.36, SD =.930), t(787.058) = −9.872, p < 0.001. Because the perceived measures are more fine-grained, we have performed our analysis with the perceived measures.

Analyses and Results We used stepwise linear regression analyses to test the model in Figure 7.1 because the correlations between homophily and tie strength, both interpersonally (r = .651) and person-to-site (r = .774), exceeded .6. The stepwise regression procedure improves the reliability of the estimation results that may otherwise be jeopardized by multicollinearity issues. We conducted a series of stepwise regressions to test, in order, (1) H1 to H3, (2) H4 to H6, and (3) H7. We conducted separate analyses for the two dependent variables: PWOM and MWI. To test H1–H3, we conducted a stepwise analysis in which the independent variables were perceived sWOM valence, the three perceived interpersonal relationship variables, and the interactions between perceived sWOM valence and the three interpersonal relationship variables.

Table 7.3. Standardized Regression Weights for Movie Watch Intention with only interpersonal relational characteristics

Model 1 Model 2 Model 3 Model 4 sWOM valence .340*** .322*** .311*** .306*** 7 Perceived interpersonal homophily Perceived interpersonal tie strength .222*** .225*** Perceived interpersonal source credibility sWOM valence *Perceived interpersonal homophily .236*** .228*** .162*** sWOM valence *Perceived interpersonal tie strength sWOM valence *Perceived interpersonal source credibility .153*** R² .115 .171 .220 .239 R² change .115 .055 .049 .019 F change 104.100 53.243 50.308 19.662 Sig. F Change <.001 <.001 <.001 <.001 Notes: ***p≤.001, ** p≤.010, *p≤.050 Tables 7.3. and 7.4. show the variables and their regression coefficients in each step. We discuss the results of the final selected models (Model 4, Table 7.3. and Model 5, Table 7.4.). In line with previous research, the effect of perceived sWOM valence on both MWI (β = .306, p < 0.001) and PWOM (β =

.241, p < 0.001) is positive and significant. In H1, we hypothesize that this effect would be reinforced by interpersonal homophily. The results show a positive interaction effect between sWOM valence and

183 Chapter 7: The impact of relational characteristics interpersonal homophily on both MWI (β = .162, p < 0.001) and PWOM (β = .109, p = 0.002). H1 is therefore confirmed. The interaction between sWOM valence and interpersonal tie strength was NOT present in the results of our stepwise linear regression and therefore not significant for MWI or for

PWOM. H2 is not confirmed. H3 hypothesized a positive interaction effect between sWOM valence and interpersonal source credibility; this interaction was positive and significant for both MWI (β = .153, p

< 0.001) and PWOM (β = .236, p < 0.001).

Table 7.4. Standardized Regression Weights for WOM intention with only interpersonal relational characteristics

Model 1 Model 2 Model 3 Model 4 Model 5 sWOM valence .258*** .250*** .246*** .241*** Perceived interpersonal homophily Perceived interpersonal tie strength .162*** .120** .114** Perceived interpersonal source credibility .109** .114** sWOM valence *Perceived interpersonal homophily .109** sWOM valence *Perceived interpersonal tie strength sWOM valence *Perceived interpersonal source .299*** .281*** .281*** .284*** .236*** credibility R² .089 .156 .182 .192 .201 R² change .089 .066 .026 .010 .010 F change 78.199 62.770 25.585 9.921 9.529 Sig. F Change <.001 <.001 <.001 .002 .002 Notes: ***p≤.001, ** p≤.010, *p≤.050

Next, to test H4–H6, we conducted two stepwise regression analyses with perceived sWOM valence, the three perceived person-to-site relationship variables, and the interactions between perceived sWOM valence and the three person-to-site relationship variables as independent variables (Tables 7.5. and

7.6.).

Table 7.5. Standardized Regression Weights for Movie Watch Intention with only person- to-site relational characteristics

Model 1 Model 2 Model 3 Model 4 Model 5 sWOM valence .340*** .338*** .335*** .331*** .333*** Perceived person-to-site homophily .155*** .159*** .107** .108** Perceived person-to-site tie strength Perceived person-to-site source credibility .093* .094* sWOM valence *Perceived person-to-site homophily .076* sWOM valence *Perceived person-to-site tie strength sWOM valence *Perceived person-to-site source .122*** .125*** .083* credibility R² .115 .139 .154 .160 .164 R² change .115 .024 .015 .006 .004 F change 104.100 22.281 14.109 5.697 3.901 Sig. F Change <.001 <.001 <.001 .017 .049 Notes: ***p≤.001, ** p≤.010, *p≤.050

In H4, we hypothesized a positive interaction effect between sWOM valence and person-to-site homophily. We indeed find this interaction for MWI (β = .076, p = 0.049) but not for PWOM. Only H4a

184 Chapter 7: The impact of relational characteristics is confirmed. Next, we expected that person-to-site tie strength would reinforce the effect of sWOM valence. We do not find an interaction effect of sWOM valence and person-to-site tie strength on MWI or on PWOM. H5 is not confirmed. The interaction effect between sWOM valence and person-to-site source credibility was positive and significant for both MWI (β = .083, p = 0.032) and PWOM (β = .092, p = 0.019), confirming H6.

Table 7.6. Standardized Regression Weights for WOM intention with only person-to-site relational characteristics Model 1 Model 2 Model 3 Model 4 sWOM valence .278*** .275*** .278*** .275*** Perceived person-to-site homophily .197*** .201*** .203*** Perceived person-to-site tie strength Perceived person-to-site source credibility sWOM valence *Perceived person-to-site homophily .057*** .098* sWOM valence *Perceived person-to-site tie strength sWOM valence *Perceived person-to-site source credibility .092* R² .077 .116 .135 .139 R² change .077 .039 .022 .006 F change 66.757 34.899 20.337 5.558 Sig. F Change <.001 <.001 <.001 .019 Notes: ***p≤.001, ** p≤.010, *p≤.050

Finally, to test H7, we included perceived sWOM valence, the three perceived interpersonal relationship variables, the three perceived person- to-site relationship variables, and the interactions between perceived sWOM valence and each of the six relationship variables as independent variables (Tables

7.7. and 7.8.).

Table 7.7. Standardized Regression Weights for Movie Watch Intention 7 Model 1 Model 2 Model 3 Model 4 Model 5 sWOM valence .340*** .322*** .311*** .306*** .306*** Perceived interpersonal homophily Perceived interpersonal tie strength .222*** .225*** .196*** Perceived interpersonal source credibility sWOM valence *Perceived interpersonal homophily .236*** .228*** .162*** .161*** sWOM valence *Perceived interpersonal tie strength sWOM valence *Perceived interpersonal source .153*** .155*** credibility Perceived person-to-site homophily .099** Perceived person-to-site tie strength Perceived person-to-site source credibility sWOM valence *Perceived person-to-site homophily sWOM valence *Perceived person-to-site tie strength sWOM valence *Perceived person-to-site source credibility R² .115 .171 .220 .239 .243 R² change .115 .055 .049 .019 .009 F change 104.100 53.243 50.308 19.662 9.527 Sig. F Change <.001 <.001 <.001 <.001 .002 Notes: ***p≤.001, ** p≤.010, *p≤.050

185 Chapter 7: The impact of relational characteristics

In H7, we hypothesized that the moderating impact of the person-to-site variables would be weaker than that of the interpersonal variables. The final model (Model 5 for MWI and Model 6 for PWOM) includes the interaction between sWOM valence and interpersonal homophily (MWI: β =.161, p < 0.001;

PWOM: β =.107, p = 0.002) and the interaction between sWOM valence and interpersonal source credibility (MWI: β =.155, p < 0.001; PWOM, β = .239, p < 0.001), in line with our conclusions for H1 and H3. None of the interactions between sWOM valence and any three of the person-to-site relational characteristics were retained, meaning that they were not significant. Thus, while two of three interpersonal relational characteristics exert a highly significant moderating effect, none of the person- to-site relational characteristics do. This is a clear indication to accept H7.

Table 7.8. Standardized Regression Weights for WOM intention

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 sWOM valence .258*** .256*** .250*** .246*** .242*** Perceived interpersonal homophily Perceived interpersonal tie strength .114** .110** .081* Perceived interpersonal source credibility .086* sWOM valence *Perceived interpersonal .103** .107** homophily sWOM valence *Perceived interpersonal tie strength sWOM valence *Perceived interpersonal .299*** .281*** .284*** .284*** .239*** .239*** source credibility Perceived person-to-site homophily .201*** .168*** .168*** .153*** Perceived person-to-site tie strength Perceived person-to-site source credibility sWOM valence *Perceived person-to-site homophily sWOM valence *Perceived person-to-site tie strength sWOM valence *Perceived person-to-site source credibility R² .089 .156 .196 .208 .216 .222 R² change .089 .066 .040 .012 .008 .006 F change 78.199 62.770 40.055 11.838 8.622 6.133 Sig. F Change <.001 <.001 <.001 .001 .003 .013 Notes: ***p≤.001, ** p≤.010, *p≤.050 Discussion

The present study shows that when considering interpersonal and person- to-site relational characteristics separately, they indeed both moderate the effect of sWOM valence on consumer responses. More specifically, for both interpersonal and person-to-site characteristics, homophily and source credibility that moderate the effect of sWOM valence on MWI and WOM intention. Although interpersonal relationships with SNS connections are not as concrete as real-life relationships, they do have an impact on how people respond to word of mouth in online social networks. Both interpersonal

186 Chapter 7: The impact of relational characteristics homophily and source credibility moderate the effect of sWOM valence on sWOM responses. The more that receivers perceive that the individual sender is like themselves, and the more credible the sender, the more positive the impact of a positive recommendation and the more negative the impact of a negative recommendation on the receivers’ behavioral intention. These effects extend the findings of previous WOM research in both offline (e.g. Buda & Zhang, 2000) and online contexts (e.g. Aghakhani et al., 2018; Ayeh, Au, & Law, 2013; Pentina et al., 2018) contexts that homophily and source credibility impact WOM responses by showing that they actually moderate the effect of WOM valence.

In contrast to our expectations, we did not find a significant moderating effect of tie strength (neither interpersonal nor person-to-site). This might be due to the correlation between homophily and tie strength. Aghakhani et al. (2018) also did not find an effect of tie strength (on sWOM readers’ cognitive attitude). They argued that consumers accept eWOM from their close ties because they perceive them as a credible source of eWOM. In other words, tie strength may have an indirect effect by influencing perceived source credibility. In our study, this explanation is less likely, as both factors were manipulated independently and the correlation between tie strength and source credibility is less than .4. The lack of effect of tie strength could also be explained by the fact that we used an unfamiliar “brand” (or movie, in this case). Bitter and Grabner-Kräuter (2016) also did not find an interaction between valence and tie strength on visiting intentions for a restaurant with which review readers were not familiar (where it was significant when review readers were familiar with the reviewed restaurant). 7

Conclusion Our core research question was to examine the claim made by Brown et al. (2007) that person-to-site relational characteristics significantly explain the responses of consumers to sWOM. Kim, Kandampully, et al. (2018) offered a partial test of this claim by studying the effects of person-to-site homophily, tie strength, and source credibility on eWOM effectiveness without taking other relevant variables, such as interpersonal relational characteristics, into account. By concurrently entering interpersonal and person- to-site relations, we believe to have provided a more accurate test of Brown et al. (2007)’s proposition.

Our findings provide no evidence for their proposition. On the contrary, our findings indicate that, when concurrently testing the moderating effect of interpersonal and the person-to-site relational characteristics, the moderating effects of the person-to-site relational characteristics are no longer

187 Chapter 7: The impact of relational characteristics significant. Consequently, the person-to-site relational characteristics do not outweigh the interpersonal relational characteristics in an SNS context. Rather, it seems to be the other way around: The interpersonal relational characteristics, and in particular homophily and source credibility, moderate the relationship between sWOM valence and our dependents, MWI and WOM intention. Even though it should be acknowledged that person-to-site homophily still had a significantly positive main effect on sWOM response, the lack of interactions is most important. We believe that this results from the fact that in SNSs, a sender is highly identifiable, compared to other online communities such as review sites.

Therefore, the sender can be held accountable for his or her own messages, and receivers’ relations with the platform, which serves only as a medium, do not add additional arguments to account for their responses (Balaji et al., 2016).

Our results also provide insights for practitioners. It is generally acknowledged that eWOM influences box office revenues of movies (Liu, 2006) and sales (Rui et al., 2013). The reason for this impact is twofold. First, as movies reach a broad audience, it can be expected that WOM impacts this audience.

Consequently, WOM will promote awareness of and interest in a movie (Liu, 2006). Second, when it is difficult to evaluate alternatives before purchase, which is the case for movies, consumers will use WOM to get more information about the product (Rogers, 1995). Research by Liu (2006) and Rui et al. (2013) indicates that pre-release, eWOM can be used to forecast box office sales. The authors further suggested that eWOM is a complementary source of information rather than a substitute, thus existing next to other types of marketing information. Our research findings indicate that it is the interpersonal relational characteristics, and primarily perceived homophily and source credibility, that moderate the effect of sWOM valence on PWOM and MWI. Yang et al. (2013) already suggested that practitioners should identify interpersonal relationships to “influence the user, encouraging participation and improving the performance of social commerce activities” (p. 74). Our research implies that marketers should try to stimulate sWOM from credible sources that are homophilous to the target audience, as these relationships reinforce the positive impact of sWOM valence on behavioral intentions. To identify credible sources marketers could look for those users who are not primarily aiming for commercial collaborations as they will be perceived as less authentic as this authenticity is the strength of sWOM and influencer marketing (De Veirman, Cauberghe, & Hudders, 2017). Moreover, it is also important to identify those consumers who have experience with the product category. For example, a tech-company could

188 Chapter 7: The impact of relational characteristics collaborate with a tech-blogger, but a collaboration with an influencer who talks about travels will be less credible. The relationship of the sWOM receiver with the site on which the sWOM appears seems less important, but it does not hurt to consider this anyway. Our findings suggest that a more homophilous Website could lead to a higher movie watch intention as well as WOM intention. Therefore, managers could try to stimulate sWOM, especially on sites that are true “matches” with the interests of their target audience.

The relationship of the sWOM receiver with the site on which the sWOM appears seems less important, but it does not hurt to consider this anyway. Our findings suggest that a more homophilous Website could lead to a higher movie watch intention as well as WOM intention. Therefore, managers could try to stimulate sWOM, especially on sites that are true “matches” with the interests of their target audience.

Limitations and Suggestions for Future Research

Limitations of the present study provide opportunities for future research. First, Chu and Kim (2011) suggested that, next to interpersonal homophily, tie strength, and trust (which is a dimension of source credibility), sWOM behavior such as opinion seeking and passing is also influenced by normative and informational interpersonal influence. We have not included these concepts in our framework because our main objective was to test the propositions of Brown et al. (2007) and the empirical findings of Kim, Kandampully, et al. (2018), who have provided clear guidelines on what constitutes homophily, tie strength, and source credibility in online social networks. Future research should explore the extent to 7 which other relationship variables that up to now have always been studied interpersonally can be translated to person-to-site equivalents and test their influence on sWOM responses as well. To date, theoretical work on the mechanisms through which person-to-site characteristics might impact consumer responses to messages on social media sites is lacking. Future research should further investigate these underlying mechanisms. Moreover, there is a need to study how to measure person- to-site relational characteristics in a reliable and valid way. In addition, the operationalization of interpersonal relational characteristics should be further explored. For example, based on research by

Fogués et al. (2014) and Tuna et al. (2016), we used the number of common friends to manipulate tie strength. However, it has also been used to operationalize homophily (Hanneman & Riddle, 2005). While our manipulation checks confirm that our manipulations were successful, the potential overlap between

189 Chapter 7: The impact of relational characteristics homophily and tie strength (due to the number of common friends) could explain the correlation between interpersonal homophily and tie strength in this study.

Future research should also explore other SNSs than Facebook, as Facebook typically contains a lot of

(inter)personal information on which users can base their evaluations of their relationship with the sender. In other online contexts, such as review sites, this information is more often lacking, and users of these sites might, therefore, appear to be more inclined to evaluate their relationship with the Website

(Kim, Kandampully, et al., 2018), which may influence the relative importance of interpersonal versus person-to-site variables. Future research should, therefore, examine whether our findings hold in other online social networks and whether the framework can and should be expanded.

Second, we use only one type of buying motivation: Movies can be categorized as hedonic products.

People use hedonic products for their aesthetic or sensory experience, for amusement, fantasy, and fun. These products are evaluated on subjective characteristics, such as shape, taste, or look (De Keyzer et al., 2017). Compared to utilitarian products, hedonic products might be more congruent with the specific use of SNSs, as they are primarily used to pass time and for amusement (Ku et al., 2013).

Moreover, the relative importance of the individual relational characteristics might differ depending on the buying motivation. For a more utilitarian product, interpersonal homophily might be less important whereas source credibility could be more important. Future research should also include more utilitarian products, such as nonfiction books and personal computers because these are more cognitively evaluated (De Keyzer et al., 2017).

The role of other boundary conditions should be further explored. First, product category involvement and prior product knowledge might impact our findings: More involved consumers or consumers with more prior knowledge about the product might process sWOM messages more centrally, which might affect the relative importance of the relational variables under study (Doh & Hwang, 2009). As mentioned, Bitter and Grabner-Kräuter (2016) found that in case a respondent is unfamiliar with the brand, tie strength with the sWOM sender or the interaction between sWOM valence and tie strength do not impact visiting intentions. They explained this effect by suggesting that consumers need prior knowledge about the brand, as well as a prior positive brand attitude before a Facebook message can impact visiting intentions. In our study, we tried to rule out potentially confounding effects of brand

190 Chapter 7: The impact of relational characteristics associations and preexisting positive or negative beliefs, attitudes, or feeling, which is in line with suggestions by Geuens and De Pelsmacker (2017). Therefore, we used an unfamiliar movie. Nevertheless, the title of the movie, the trailer of the movie might have induced (un)favorable attitudes towards the movie regardless of the relationship with the sWOM sender or the social networking site.

Future research could use either many existing brands to neutralize the confounding effects or a control group to examine the effect of familiar brands (Geuens & De Pelsmacker, 2017).

Next, it might be interesting to examine users’ motivations to use the SNS as a potential moderator. In general, three motivations to use a medium can be distinguished: pass time, amusement, and information seeking. Research has indicated that SNSs are primarily used to pass time and for amusement (Ku et al., 2013). Users who mainly use SNSs for these motivations are less likely to assess the aspects of the relationship with the source (both personal and with the site) compared to users who use SNSs for information seeking (Metzger, 2007). Our research did not take user motivations into account. Therefore, future research could examine the role of users’ motivation to use SNS in a real-life setting. Also, as mentioned in the Discussion section, the context in which our sWOM message was placed contained a lot of interpersonal information. This might explain the missing impact of person-to- site relational characteristics. Future research should study the model at hand in other online contexts, such as review sites or other social media in which fewer interpersonal relational characteristics are present. 7

Moreover, personality traits might also impact the (relative) effects of the relational antecedents. For example, previous research has indicated that susceptibility to interpersonal influence might have a strong impact on consumers responses to eWOM messages (Chu & Kim, 2011). Users who are more susceptible to interpersonal influences are more likely to comply with eWOM messages overall (Chu &

Kim, 2011). When they do so, they may be more likely to disregard their perceptions of interpersonal homophily, tie strength, and source credibility.

Finally, we measured self-reported behavioral intentions. Future research could examine real social network data, user log records, or other data sources. There are a number of measures that can be applied to measure homophily and tie strength (e.g., mutual friends, last interaction; e.g. Gilbert &

Karahalios, 2009). This would strengthen theoretical development in the field of SNSs.

191

General Conclusion

8 Chapter 8: General Conclusion

This chapter first describes a summary of the main findings reported in the six empirical chapters of the dissertation. Next, it describes the contribution of the present work to both theory and practice. Finally, it discusses a number of limitations of the studies and elaborates on suggestions for future research.

Main Findings

The overall goal of the dissertation was to explore consumers’ responses to brand communication on social media. By studying both personalized advertising and word-of-mouth on social networking sites, the dissertation combines insights into one-way (personalized advertising) and two-way communication

(sWOM) between consumers and brands. By exploring both the processing mechanisms and boundary conditions of brand communication on SNSs, we provide a better understanding of how and under what circumstances brand communication on SNSs affect consumers’ responses.

The first goal of the dissertation was to determine the relative importance of different personalization elements to elicit perceived personalization. Previous research had already indicated that actual and perceived personalization do no automatically match (Kramer et al., 2007; Li,

2016). The study in chapter 2 indicates that each of the elements under study (interests, location, age, life events, gender and friend referrals) are able to induce perceived personalization. Interests seem to be the most important. These results also remain stable across product characteristics and demographics. Indeed, prior studies using interests as personalization elements have reported successful manipulations of personalization (Kalyanaraman & Sundar, 2006; Li, 2016).

The second objective was to explain the positive and negative effects of personalized advertising on consumers’ attitudes and behavioral intentions. The dissertation fills the existing knowledge gap concerning the processing mechanisms of personalized advertising on social networking sites. As mentioned by Boerman et al. (2017), these mechanisms are not yet clearly understood.

Chapters 3 and 4 show a strong positive indirect effect on brand attitude, click intention and positive word-of-mouth intention through personal relevance. This is in line with previous research identifying personal relevance as the most important mediator in the relationship between personalized messages and consumers’ responses (e.g., Aguirre et al., 2015; Bleier & Eisenbeiss, 2015; Kim & Huh, 2017). In

Chapter 4, perceived intrusiveness and perceived entertainment were added as a first-order mediator, next to personal relevance. Contrary to our expectations, a positive indirect effect of advertising

194 Chapter 8: General Conclusion personalization via perceived intrusiveness on brand attitude, click intention and positive word-of-mouth intention was found. Based on van Doorn and Hoekstra (2013) and White et al. (2008), it was expected that using more personal information would increase perceived intrusiveness. However, the findings indicate that the opposite is true. This might be that due to the fact that because the information in the advertisement is more valuable, the advertisement is perceived as less intrusive compared to the non- personalized advertisement (e.g., Ketelaar et al., 2018; Kim & Han, 2014).

Perceived entertainment, in addition, was found to be a positive first-order mediator in the relationship between perceived personalization and brand attitude, click intention and positive word-of-mouth intention. This is in line with previous research. Kim and Han (2014) already posited that a personalized message is more likely to reflect a consumers’ own needs and, therefore, will increase perceived entertainment. Edwards et al. (2013) also found that more entertaining advertisements are less likely avoided by consumers. Chapter 4 also added two second-order mediating variables: self-brand connection and reactance to the advertisement. The results indicate that self-brand connection is indeed a second-order mediator in the relationship between perceived personalization and brand attitude, click intention and positive word-of-mouth intention. More specifically, perceived personalization has a positive effect on self-brand connection via perceived personal relevance and perceived entertainment, but not via perceived intrusiveness. Therefore, the positive indirect effect of perceived personalization through perceived intrusiveness cannot be explained by self-brand connection. The positive effects of perceived personalization on self-brand connection can be explained via the self-referencing process: both via perceived relevance and perceived entertainment, a consumer’s own psychological needs can 8 be addressed. However, this is not the case for perceived intrusiveness. In turn, self-brand connection positively affects brand attitude, click intention and positive word-of-mouth intention. Our findings also indicate that perceived personalization can decrease reactance to the advertisement via perceived entertainment, perceived relevance and perceived intrusiveness. Edwards et al. (2013) also find that perceived entertainment can decrease the psychological reactance toward a personalized advertisement due to the value that it generates for consumers. Contrary to our expectations, reactance to the advertisement did not significantly affect brand responses. This might be explained by the findings of

Marotta, Zhang, and Acquisti (2015), who noted that while consumers do not like being tracked online,

195 Chapter 8: General Conclusion they often appreciate the benefits of online targeted advertising. As a result, they will not be inclined to try to react against the advertisement.

Next, Chapter 5 focused on the relative importance of positive (benefits) and negative (costs) effects of perceived personalization. The findings indicate that the (expected) positive effect via personal relevance is outweighed by perceived creepiness. Even using only one personalization element (friends’ referrals) is enough to creep consumers out, in which case personal relevance does not have a significant effect.

The third objective of the dissertation was to examine boundary conditions of the effects of personalized advertising on consumers’ attitudes and behavioral intentions, and the associated processing mechanisms. Chapters 3 and 4 examined the impact of the attitude toward the social networking site. Chapter 3 examined its impact on the direct relationship between perceived personalization on brand attitude and click intention. The results indicate that, for a high involvement product, attitude toward the social networking site does not moderate the direct relationship. However, it does moderate the effect of perceived personalization on click intention (but not on brand attitude) for a low involvement product. The positive effect of perceived personalization is only significant for consumers with a moderate to positive attitude toward the social networking site. The difference between the high and low involvement product might be due to a ceiling effect: the effect size of the main effect of perceived personalization is much larger for the high involvement product than for the low involvement product. Perhaps personalization fits better with a high involvement product, compared to a low involvement product. Furthermore, the motivation to process the personalized advertisement for the high involvement product might already be so high that a positive attitude toward the social networking site cannot further increase the positive effect of perceived personalization. Next, Chapter

4 examined whether the attitude toward the social networking site moderates the indirect effects of personal relevance and perceived intrusiveness on consumers’ responses. No significant interaction effects on the indirect paths were found. Even though the focus of the study was on the moderation of the indirect effect is, the findings indicate a moderation of the direct effect of perceived personalization on positive word-of-mouth intention and click intention: for those with a negative attitude toward the

196 Chapter 8: General Conclusion social networking site, the effect of perceived personalization on positive word-of-mouth and click intention decreases, whereas for those with a moderate or positive attitude, the effect increases.

Chapter 4 also examined the moderating role of perceived privacy protection by the social networking site. Our findings indicate that the indirect effect via perceived relevance on click-intention (but not on brand attitude or positive word-of-mouth) is indeed moderated by perceived privacy protection by the social networking site. More specifically, when the level of perceived privacy protection by the social networking site increases, the indirect effect of perceived personalization on click intention via personal relevance becomes stronger. One explanation might be that trust, which can be created by being clear about privacy protection measures, can spill over from a website to the embedded, personalized, advertisements (Aguirre et al., 2015). Moreover, Mpinganjira and Maduku (2019) show that perceived privacy control has a negative influence on privacy concerns and, indirectly, on ad avoidance. When users are confident their privacy is protected, they may appreciate personalized advertising more. As a result, they might value the personal relevance of the personalized advertisement more which results in positive brand responses.

Chapter 5 studied the moderating role of the type of source (i.e., the advertiser). The negative impact of personalization through perceived creepiness only occurred for health websites, governmental websites and online newspapers, but not for commercial websites. This confirms the expectation, based on Acquisti et al. (2015) and Nissenbaum (2010), that the way in which people weigh the benefits and the costs of personalized advertisements is source-dependent. When the stakes are higher, when people 8 might be less accustomed to personalized advertising, the risks are more important than when the information used in the personalized advertisement is less personal or when people might be more accustomed to personalized advertisements.

A fourth, and final, objective of the dissertation was to examine boundary conditions of the message valence effect of sWOM messages on consumers’ attitudes and behavioral intentions. Chapter 6 examined the moderating effect of tone of voice (factual vs emotional) and service type (utilitarian vs hedonic). Our findings indicate that message valence exerts a greater impact on attitude, purchase intention and positive WOM intention when the message uses a factual, compared to an emotional, tone of voice. Message valence also has a stronger effect on consumer responses when

197 Chapter 8: General Conclusion evaluating an sWOM message about a hedonic service compared to a utilitarian service. Finally, we expected that, compared to mismatching, matching the tone of voice to the type of service would have led to a greater impact of message valence on consumer responses. However, in our study, for both utilitarian and hedonic services, the informative effect of a factual message outweighs the potential matching effect in terms of tone of voice.

Chapter 7 examined the moderating role of interpersonal and person-to-site relational characteristics on the effect of message valence on behavioral and positive word-of-mouth intention. Our findings indicate that interpersonal homophily and source credibility strengthen the effect of sWOM valence on behavioral and positive word-of-mouth intention. When only considering person-to-site relational characteristics as moderating variables, person-to-site homophily and source credibility also strengthen the effect of sWOM valence. However, when considering both the interpersonal and person-to-site relational characteristics, the effects of all person-to-site relational characteristics are non-significant.

Theoretical contributions

The findings of this dissertation extend prior research in the field of brand communication on social media by studying both personalized advertising and word-of-mouth on social networking sites.

The findings of the Chapters 2 - 5 help shed light on the findings of previous studies (e.g., Maslowska,

Smit, et al., 2011) who did not find a significant effect of personalization on attention to the message, attitudes towards the message and the brand or patronage intentions. On the one hand, a number of studies used actual personalization as a 0/1 variable, whereas the broader picture might be more nuanced. The actual personalization (how it is manipulated by researchers or advertisers) might be perceived as generic by some (Kramer et al., 2007). As such, using actual personalization as input might not be sufficient. For example, Chapter 2 shows that even though all personalization elements did increase the perception of personalization, interests are the most important personalization element to elicit perceived personalization. In turn, friends’ referrals, life events, and gender are the least important.

Indeed, the pretest in Chapter 5 indicated that using only friends’ referrals leads to a low level of personalization. When friends’ referrals are combined with gender, or with both gender and interests, the perception of perceived personalization increases. In Chapters 3 and 4, perceived personalization was used. These manipulations led to at least a certain degree of variance in the perceived

198 Chapter 8: General Conclusion personalization score of the advertisements. Therefore, we recommend future researchers to use perceived personalization either as a manipulation check or to use perceived personalization in their analyses.

Moreover, personal relevance is an important variable in explaining the effects of personalized advertising. However, when consumers are creeped out by the personalized advertisement, the positive effect via personal relevance is overruled. Another important finding of the dissertation is that personalization does not increase the intrusiveness of an advertisement as was found by White et al. (2008) and van Doorn and Hoekstra (2013). On the contrary: personalization, compared to non- personalization, seems to provide added value for the consumer, therefore decreasing intrusiveness of the advertisement. Finally, the findings of the current dissertation show that perceived entertainment is also a relevant mediating variable. By providing personally relevant and entertaining advertisements, personalized advertising is able to address the psychological needs of consumers. As such, both personal relevance and perceived entertainment increase the likelihood that a consumer will connect his own self-image to the brand image (self-brand connection). This self-brand connection, in turn, improves consumer responses such as brand attitude, click intention and positive word-of-mouth intention. Even though perceived personalization has a positive effect on reactance to the advertisement via perceived intrusiveness and a negative effect via perceived entertainment and personal relevance, reactance to the advertisement does not seem to affect consumer responses.

Based on the privacy calculus theory, we examined whether the benefits (e.g. perceived relevance or 8 perceived entertainment) of personalized advertising in social networking sites would be better in explaining the personalization effect on brand attitude, click intention and positive word-of-mouth intention, compared to the costs or the risks of personalized advertising (e.g. perceived intrusiveness or perceived creepiness). Our results suggest that the benefits of personalization can outweigh the costs, except when consumers are creeped out by the personalization, in which case the costs outweigh the benefits. This could be explained by the fact that by providing consumers with personalized information the advertisers is able to bring added value to the consumer: it is more relevant, more entertaining and more in line with the consumers own goals, thus less intrusive. However, this added value of these advertisements might not always outweigh the affective response of creepiness. This is

199 Chapter 8: General Conclusion especially true for non-commercial advertisers. As shown in Chapter 5, the effect of creepiness only occurred for non-commercial advertisers. This might be explained by the fact that consumers learn which kind of personal information is appropriate to use in advertising contexts, and which kind of personal information is not appropriate to use (Nissenbaum, 2010). Moreover, in line with the Persuasion

Knowledge Model (Friestad & Wright, 1994), it can be expected that consumers learn that commercial sources use personalized advertising as a persuasive tactic, whereas this might not be the case for non- commercial sources. As a result, consumers might be more inclined to resist personalized advertising from non-commercial sources, compared to commercial source from which it is clear that they have a persuasive intent. Our findings suggest that the privacy calculus theory does not only apply to self- disclosure behavior (Culnan & Armstrong, 1999; Dienlin & Metzger, 2016; Laufer & Wolfe, 1977) but can also be applied to explain how consumers respond to advertisers using that self-disclosed information (Demmers et al., 2018).

Our findings indicate that the attitude toward the social networking site and perceived privacy protection by the social networking site are important boundary conditions of the effects of personalized advertising. The attitude toward the social networking site appears to play a role on the direct effect of perceived personalization and the indirect effect via personal relevance on click intention, but not on the indirect effects via personal relevance or perceived intrusiveness on brand attitude or positive WOM intention. In Chapter 3 we established the moderating role on the direct effect, however, theoretically it is more interesting to examine the moderation of the indirect effect. Perceived privacy protection by the social networking site does affect the indirect relationship between perceived personalization and click intention via personal relevance. The type of advertisers is also an important moderating variable: our findings suggest that, for a commercial site, there appears to be no negative effect of perceived personalization on brand attitude through perceived creepiness. In other contexts (health, governmental and news), higher levels of personalization trigger the perception of creepiness which in turn negatively reflects on brand attitude. Apparently, the use of personal information is appropriate in some contexts, but not in others (Acquisti et al., 2015).

Finally, the findings of the second part of the dissertation contribute to the understanding of the message valence effect of sWOM messages. Previous research demonstrated that positive eWOM

200 Chapter 8: General Conclusion

(compared to negative eWOM) leads to more positive consumer responses (e.g., Bae & Lee, 2011;

Purnawirawan et al., 2015). In response to a call from Barger et al. (2016) to study the effects of content and product factors, we examined the effects of the tone of voice of an sWOM message and of service type on the relationship between message valence and consumer responses. Our findings suggest that a factual tone of voice, compared to an emotional one, strengthens the effect of message valence on brand attitude, purchase intention and WOM intention. This finding is in line with previous research in online reviews (e.g., Willemsen et al., 2011). This can be explained by the accessibility- diagnosticity model (Feldman & Lynch, 1988): compared to emotional information, factual information is more diagnostic because it enables the consumer to asses the quality and the performance of the product or service in the sWOM message. Moreover, the attribution theory (Kelley, 1973) suggests that when the information in an sWOM message can be attributed to the product, compared to the person or the circumstances, that information becomes more important in the consumer’s evaluation of the product or service discussed in the sWOM message. This effect is present regardless of the product or service in the sWOM message: our findings suggest that matching the tone of voice to the sWOM message does not strengthen the effect of sWOM valence on a consumer’s responses. Klein and Melnyk (2016) have already suggested that matching (i.e., using a factual tone of voice for a utilitarian service and an emotional tone of voice for a hedonic service) matters less for hedonic services, due to the fact that arguments for this type of service might not be scrutinized in detail. They suggest that the presence of arguments offers a heuristic cue on which evaluations can be based. However, a message for a hedonic service, compared to a utilitarian one, also reinforces the effect of message valence. This effect 8 was also previously found in other online contexts (e.g., Park & Lee, 2009; Park & Lee, 2008; Wu &

Wang, 2011) and thus also holds in the context of social networking sites. This might be explained by the fact that hedonic services are harder to evaluate, compared to utilitarian ones. As a result, the impact of sWOM valence on consumers’ responses might be stronger for hedonic services, especially when the message is factual.

Brown et al. (2007) claimed that person-to-site relational characteristics are important to explain the responses to sWOM messages. Kim, Kandampully, et al. (2018) partially tested this claim by studying the effects of homophily, tie strength and source credibility of a site without considering other relevant variables such as interpersonal relational characteristics. We believe that we provided a more accurate

201 Chapter 8: General Conclusion test of Brown, Broderick and Lee’s (2007) proposition. Our results do not provide evidence for their claim. On the contrary, our findings suggest that, when taking both interpersonal and the person-to- site relational characteristics into account, only the interpersonal characteristics strengthen the effect of message valence on consumer responses. In that situation, the person-to-site relational characteristics are irrelevant. We believe that this stems from the fact that in social networking sites, the senders of sWOM messages are highly identifiable, compared to other online communities such as review sites.

Therefore, the sender can be held accountable for his own messages (Balaji et al., 2016).

Managerial contributions

In sum, using social media for brand communication, both in terms of personalized advertising and word-of-mouth, can have positive effects on brand attitude, click intention and positive word-of-mouth intention. The findings of the present dissertation have a number of implications for managers. The following outlines a number of the elements and conditions which should be taken into account when managers wish to use social media in their brand communication mix.

First, our findings provide guidelines in which personalization elements should be used in personalized advertisements to elicit perceived personalization. Consumers’ interests seem to be the most effective to evoke perceptions of personalization, whereas friends’ referrals, life events, and gender seem to be least effective. Nevertheless, all personalization elements contribute to the perception of personalization and can therefore be combined to elicit the highest level of personalization possible. However, managers should consider both the costs and the benefits of this: combining all personalization elements could increase the cost of the advertisement and it might not be cost-effective to pay more for elements that contribute less to perceived personalization.

Moreover, our findings indicate that perceived personalization can improve personal relevance, perceived entertainment and can even decrease the level of perceived intrusiveness of the advertisement. At the same time, levels of personalization that become too high are perceived as creepy and, in that case, the effect of personal relevance can be overruled by the feeling of creepiness, decreasing brand attitudes. It is thus very important for managers to think about the personalization elements to be used in their personalized advertisements. More specifically, the findings in Chapter 2 suggest that consumers find the use of location, life events and friend referrals more annoying than

202 Chapter 8: General Conclusion useful. As life events and friend referrals only weakly contribute to the perception of personalization, which is needed to affect positive outcomes such as brand attitude, click intention or positive word-of- mouth intention, we suggest to not use these two elements in personalized advertisements. Even though friends’ referrals seem to be a popular advertising technique in social media, it appears to less effective than assumed. For example, Windels et al. (2018) also suggest that the use of friends’ referrals in advertisements underperforms in comparison with advertisements that do not use friends’ referrals in terms of visual attention. To our knowledge, no previous research has used life events in their study.

However, it is clear that to increase the perceived relevance of the advertising message the life events must fit the product. For example, when advertising baby products it will be more interesting to target parents who indicated they just had a baby instead of targeting parents of teenagers.

Our findings also indicate that the type of advertiser is important for how consumers weigh the benefits and costs of personalized advertising. For a commercial website, the negative effect via creepiness does not play a role. As such, commercial advertisers can take their personalization efforts much further.

However, for other contexts (health, governmental and news) advertisers should be more careful with the level of personalization used in their advertisements. In these contexts, the negative effect of creepiness does appear, and, in that case, the use of more personalization elements might negatively impact brand attitude.

Moreover, brand managers should be aware of the social media sites they place their personalized advertising on. Our findings suggest that SNSs that are liked better and that are clear about their privacy 8 protection measures can reinforce the positive effect of perceived personalization. Moreover, when a specific platform is selected, advertisers could opt to target those consumers with a more positive attitude toward the platform. Our findings also have implications for social media platform managers.

Our findings also provide implications for social media platform owners. It is clear that it is important to create a positive brand image as a platform: by creating a positive image the platform can reinforce the positive effect of personalization via relevance on consumers attitudes and intentions. Moreover, by being clear about the privacy protection measures the platform takes, this indirect effect via relevance can also be strengthened. This is in line with Mpinganjira and Maduku (2019) who show that privacy control is an important element in the perception of ethical behavior by organizations as it decreases

203 Chapter 8: General Conclusion advertising avoidance. They posit that, contrary to common belief, ethical practices could actually increase advertising effectiveness.

With regard to word-of-mouth messages on social networking sites, our findings indicate that message valence has a consistent effect on consumer responses: more positive message lead to more positive responses. Therefore, managers should try to encourage their customers to post positive sWOM messages. Since factual messages, compared to emotional ones, reinforce this positive message valence effect, we advise sWOM senders and marketers soliciting positive sWOM messages to back their messages up with facts and to avoid an emotional tone of voice. Furthermore, our findings suggest that sWOM receivers are more influenced by the message valence effect for hedonic services, compared to utilitarian ones. This is due to the fact that hedonic services are difficult to evaluate and, therefore, consumers will rely more on WOM (or sWOM in this case) to get more information about the product

(Rogers, 1995). Therefore, marketers of hedonic services should actively invest more effort in trying to elicit positive factual sWOM and countering negative sWOM.

Our findings also suggest interpersonal relational characteristics, and primarily homophily and source credibility, moderate the effect of sWOM valence on behavioral intentions. Therefore, we suggest that marketers should try to stimulate sWOM from credible sources that are highly homophilous to the target audience, as this will reinforce the positive impact of sWOM on behavioral intentions. The relationship with the site on which sWOM messages appear seems less important – which is contrary to the findings for personalized advertisements – but of course it does not hurt to consider this anyway. Our findings suggest that a more homophilous website could lead to stronger behavioral intentions. Therefore, managers could try to stimulate sWOM on sites that match the interests of their target audience.

Limitations and directions for future research

A number of limitations of the present dissertation have already been discussed in each of the empirical chapters with respect to the individual studies. Although more specific issues will not be reiterated here, a discussion of the more important limitations, as well as ways to address them in future follows next.

The current studies use an experimental approach in which different factors (i.e., different

(combinations of) personalization elements, type of advertiser, message valence, tone of voice, type of product or relational characteristics) were systematically manipulated. Moreover, throughout the

204 Chapter 8: General Conclusion different studies, we have measured mediating variables such as perceived relevance, intrusiveness or entertainment. We study this in the context of the most popular social networking site to date, Facebook. Even though other social media platforms use similar personalization elements to personalize advertisements. Our findings already suggest that the context, i.e., the social media site in which the advertisement is placed, has a significant moderating effect on the relationship between perceived personalization and outcomes. Therefore, the same personalization element might not have the same effect on consumers’ responses in the same platform. Previous research has indicated that consumer use different social media to perform different brand-related activities (Muntiga, Moorman, & Smit, 2011). Moreover, Buzeta, De Pelsmacker, and Dens (2018) show that, in general, empowerment and remuneration are important motivations in the context of social media. However, they also discovered differences between different types of social media. Voorveld, van Noort, Muntinga, and Bronner (2018), for example, find that Facebook is primarily used for social interaction, whereas Instagram is primarily used for entertainment. Moreover, Chang, Liu, and Shen (2017) find that users’ trust differs between

LinkedIn and Facebook. They show that privacy concern has a stronger impact on users’ trust for LinkedIn than for Facebook. They explain their findings by positing that LinkedIn is used to post work- related information and this information might be seen as more confidential than information shared on

Facebook. As such, when comparing different social media platforms, it might be that the indirect effects of personalization via creepiness on consumers’ source attitude found in Chapter 5 might be moderated by the social media platform. Therefore, also in response to the call of Voorveld (2019), future research could examine how different platforms affect the relationships studied in this dissertation. 8

In the first part on personalized advertising we tested perceived relevance of the advertisement and perceived intrusiveness of the advertisement as more cognitive mediating variables and perceived entertainment and perceived creepiness as more affective mediating variables. Nevertheless, due to our methodology our research examines the processing of personalized advertising from a more cognitive, than affective, processing viewpoint. Future research could focus more on the affective processing by using more implicit methodology, for example by using eye-tracking or implicit memory tests, and by examining more potential mediating affective variables, such as social interaction, remuneration, which, then, should congruent or incongruent with users’ motivations to use social media.

205 Chapter 8: General Conclusion

In each of the studies on personalized advertising, we manipulated personalization very overtly (e.g., by using it in the slogan of the advertisement). Although this is not uncommon, personalization techniques are more often covert. Past research on personalization and privacy present mixed findings on the differences between covert and overt personalization. For example, Sundar and Marathe (2010) suggest that covert personalization induces more privacy concerns, compared to overt personalization, whereas Chen and Sundar (2018) do not find differences between overt and covert personalization in a mobile app in terms of privacy concern, attitude toward the application, behavioral intention or purchase intention. Future research could examine the difference between overt and covert personalization in social media further. Moreover, to our knowledge, there is only initial evidence on what consumers believe is personal space, social or public space on social media. Xie and Karan (2019) report that consumers set up different privacy boundaries for different types of personal information: “social identity information and daily life and entertainment information tended to be shared more freely, while personal contact information was mostly withheld” (p. 1).

The current dissertation tests the effects of personalization and word-of-mouth communication in social networking sites on consumers’ attitudes and intentions at specific moments in time, and as such, is a collection of snapshots. For example, the data for Chapter 3 were collected at a moment in time in which banner advertisements were common practice. When the data collection for Chapters 2, 4 and 5 were collected native advertisements had become common practice. Currently, more and more interactive formats of advertising are used in social networking sites (e.g., video or advertisements which can be clicked to buy one or more specific products displayed in the advertisement). Moreover, the use of these platforms is also changed throughout the years. Based on the Persuasion Knowledge

Model (PKM) (Friestad & Wright, 1994) it can be expected that social networking site users learn to cope with advertisements in their social networking sites and they can learn to recognize these advertisements. This might explain the unexpected positive indirect effect of perceived personalization on consumers’ brand responses via perceived intrusiveness: users might not feel as intruded when they come across an advertisement now compared to when advertising on social networking sites was less common. Future research could examine moderating role of persuasion knowledge as well as the longitudinal effects of perceived personalization on consumers’ brand responses.

206 Chapter 8: General Conclusion

Moreover, when manipulating friends’ referrals in Chapter 2 and 5 it was decided to use a generic phrasing (3 of your friends like [brand]) in order to increase internal validity. However, real-life Facebook advertising would state the names of specific friends. We did not want to confound our findings by allowing respondents to, for example, provide real names of friends. This might, in line with Chapter 6, influence the perception of personalization and the resulting evaluation of the message. For example, respondents providing names of friends which are highly similar (homophilous) to them might perceive the message as more personalized compared to those who provide names of friends which share less characteristics with the respondent. Agarwal et al. (2017), indeed, show that it is not the number of friends that drive effects on e.g., click-through, but the similarity (homophily) with the friends referred to by the brand or the product. Future research could examine the impact of the interpersonal relational characteristics as moderators of the relationship between (perceived) personalization and consumers’ attitudes and behavior(al intentions).

All personalized advertisements and sWOM messages were displayed in a static context. In all chapters, except for Chapters 2 and 5, respondents were exposed to an advertisement in a mimicked Facebook homepage or profile page. The lack of interactivity with the social networking site itself could have influenced the results. As with any research, there is a tradeoff between internal and external validity.

We suggest future research to perform field experiments in order to provide insights on the generalizability of our findings.

In most chapters, we examine behavioral intentions, such as click intention or positive WOM intention, 8 as the dependent variables. The Theory of Reasoned Action and the Theory of Planned Behavior (e.g.,

Ajzen, 2002) posit that attitudes and behavioral intentions are important antecedents of actual behavior, and many studies have confirmed this (e.g., Armitage & Conner, 2001; De Cannière et al., 2009).

However, this relationship is not perfect (e.g., Sheeran, 2002; Van Ittersum, 2012). We advise future research to include actual behavioral measures such as click-through or sales, which could also be achieved by conducting field experiments.

In most chapters, we used either fictitious brands or brands with which the respondents were not familiar with to avoid brand familiarity effects (Geuens & De Pelsmacker, 2017). This limits the scope of our research and the generalizability of our findings. As mentioned in Chapter 1, our studies focus

207 Chapter 8: General Conclusion primarily on the top (TOFU) and middle (MOFU) of the digital marketing funnel. Future research could test the generalizability of our findings by focusing on familiar brands. Moreover, in the bottom of the marketing funnel (BOFU) other marketing actions might more important as consumers in that stage are loyal consumers and as such, marketers need to set up actions to retain these consumers. In this stage consumers turn more to online reviews or to word-of-mouth communication on social media. This is also the phase were influencers can become important sources of information. Future research could compare the impact of brand communications on social networking sites, or social media in general, throughout the different stages of the marketing funnel.

Furthermore, when there were other messages present on the Facebook page, these were intentionally kept as neutral, yet realistic, as possible (e.g. a profile picture update) to avoid confounds. However, the mood induced by other messages might also have an impact on the processing of brand communication on sWOM. In line with the context appreciation theory, the context can also be used as a source of information in the processing mechanism. For example, a positive mood, as a result of the context, can be transferred to consumers’ responses (De Pelsmacker et al., 2002).

Finally, brands are increasingly forging alliances with influencers to promote their products (De Veirman et al., 2017). Influencer marketing is considered as more effective compared to traditional advertising tactics (de Vries, Gensler, & Leeflang, 2012). The robustness of the findings from our studies could be tested in the context of influencer marketing. On the one hand when influencers target a specific target group with their posts the findings of part one of the current dissertation seem to suggest that they can increase relevance and entertainment and decrease intrusiveness. On the other hand, based on the findings of part two it could be expected that influencers should use a more factual tone of voice when posting about brands. In line with Chapter 6, forging alliances with influencers might be more important for more hedonic brands, compared to more utilitarian brands. Finally, Chapter 7 suggests that relational characteristics are important when selecting an influencer. We advise future research to test the robustness of these findings in the context of influencer marketing.

208

References

R References

Aaker, J. L. (1997). Dimensions of Brand Personality. Journal of Marketing Research, 34(3), 347-356. doi:Doi 10.2307/3151897 Aaker, J. L., & Lee, A. Y. (2001). "I" seek pleasures and "we"avoid pains: the rol of self-regulatory goals in information processing and persuasion. Journal of Consumer Research, 28(1), 33-49. Abrahamse, W., Steg, L., Vlek, C., & Rothengatter, T. (2007). The Effect of Tailored Information, Goal Setting, and Tailored Feedback on Household Energy Use, Energy-Related Behaviors, and Behavioral Antecedents. Journal of Environmental Psychology, 27(4), 265-276. doi:DOI 10.1016/j.jenvp.2007.08.002 Acquisti, A., Brandimarte, L., & Loewenstein, G. (2015). Privacy and human behavior in the age of information. Science, 347(6221), 509-514. AdEspresso. (2016). Types of Facebook Ads. Retrieved from http://adespresso.com/academy/guides/facebook-ads-beginner/facebook-ads-types/ Agarwal, A., Lee, S.-Y., & Whinston, A. B. (2017). Word-of-Mouth in Social Media Advertising: 'Likes' on Facebook Ads. Retrieved from https://ssrn.com/abstract=3065564 Aghakhani, N., Karimi, J., & Salehan, M. (2018). A Unified Model for the Adoption of Electronic Word of Mouth on Social Network Sites: Facebook as the Exemplar. International Journal of Electronic Commerce, 22(2), 202-231. Aguirre-Rodriguez, A., Bosnjak, M., & Sirgy, M. J. (2012). Moderators of the self-congruity effect on consumer decision-making: A meta-analysis. Journal of Business Research, 65(8), 1179-1188. Aguirre, E., Mahr, D., Grewal, D., de Ruyter, K., & Wetzels, M. (2015). Unraveling the Personalization Paradox: The Effect of Information Collection and Trust-Building Strategies on Online Advertisement Effectiveness. Journal of Retailing, 91(1), 34-49. Ahluwalia, R., & Gürhan-Canli, Z. (2000). The effects of extensions on the family brand name: An accessibility-diagnosticity perspective. Journal of Consumer Research, 27(2), 371-381. Ahluwalia, R., Unnava, H. R., & Burnkrant, R. E. (2001). The Moderating Role of Commitment on the Spillover Effect of Marketing Communications. Journal of Marketing Research, 38(4), 458-470. doi:DOI 10.1509/jmkr.38.4.458.18903 Ajzen, I. (2002). Perceived Behavioral Control, Self-Efficacy, Locus of Control, and the Theory of Planned Behavior. Journal of Applied Social Psychology, 32(4), 665-683. Anand, B. N., & Shachar, R. (2009). Targeted advertising as a signal. Quantitative Marketing and Economics, 7(3), 237-266. Armitage, C., & Conner, M. (2001). Efficacy of the Theory of Planned Behaviour: A meta-analytic review. British Journal of Social Psychology, 40(4), 471-499. Arora, N., Dreze, X., Ghose, A., Hess, J. D., Iyengar, R., Jing, B., . . . Zhang, Z. J. (2008). Putting one- to-one marketing to work: Personalization, customization and choice. Marketing Letters, 19(3/4), 305-321. Asdemir, K., Kumar, N., & Jacob, V. S. (2012). Pricing Models for Online Advertising: CPM vs CPC. Information Systems Research, 23(3-part-1), 599-848. doi:https://doi.org/10.1287/isre.1110.0391 Atzmüller, C., & Steiner, P. M. (2010). Experimental Vignette Studies in Survey Research. Methodology, 6(3), 128-138. Awad, N. F., & Krishnan, M. S. (2006). The Personalization Privacy Paradox: An Empirical Evaluation of Information Transparency and the Willigness to be Profiled Online for Personalization. MIS Quarterly, 30(1), 13-28. Ayeh, J. K., Au, N., & Law, R. (2013). "Do We Believe in TripAdvisor?" Examining Credibility Perceptions and Online Traveler's Attitude toward Using User-Generated Content. Journal of Travel Research, 52(4), 437-452.

210 References

Babić Rosario, A., Sotgiu, F., De Valck, K., & Bijmolt, T. H. A. (2016). The Effect of Electronic Word of Mouth on Sales: A Meta-Analytic Review of Platform, Product, and Metric Factors. Journal of Marketing Research, 53(3), 297-318. Bachleda, C., & Berrada-Fathi, B. (2016). Is negative eWOM more influential than negative pWOM? Journal of Service Theory and Practice, 26(1), 109-132. Bae, S.-W., & Lee, J. (2011). Gender differences in consumers’ perception of online consumer reviews. Electronic Commerce Research, 11(2), 201-214. Baek, T. H., & Morimoto, M. (2012). Stay Away From Me. Journal of Advertising, 41(1), 59-76. Balaji, M. S., Khong, K. W., & Chong, A. Y. L. (2016). Determinants of negative word-of-mouth communication using social networking sites. Information & Management, 53(4), 528-540. Bang, H., & Wojdynski, B. W. (2016). Tracking users' visual attention and responses to personalized advertising based on cognitive demand. Computers in Human Behavior, 55, 867-876. Bansal, H. S., & Voyer, P. A. (2000). Word-of-Mouth Processes Within a Services Purchase Decision Context. Journal of Service Research, 3(2), 166-177. Barger, V. A., Peltier, J. W., & Schultz, D. E. (2016). Social media and consumer engagement: A review and research agenda. Journal of Research in Interactive Marketing 10(4), 268-287. Barnard, L. (2014). The Cost of Creepiness: How Online Behavioral Advertising Affects Consumer Purchase Intention. (Doctor of Philosophy in Mass Communication), University of North Carolina, Chapel Hill. Bátor, A., & Lengyel, Z. (2014). Performance Based Online Marketing in Theory and In Practice. Pannon Management Review, 3(2), 11-43. Batra, R., & Ahtola, O. T. (1991). Measuring the Hedonic and Utilitarian Sources of Consumer Attitudes. Marketing Letters, 2(2), 159-170. Bedgood, L. (2017). Life Event Marketing and How to Target Consumers When It Matters Most. Retrieved from http://www.righttimemarketer.com/2017/02/02/life-event-marketing-target- consumers-matters/ Bergkvist, L., & Rossiter, J. R. (2008). The role of ad likability in predicting an ad's campaign performance. Journal of Advertising, 37(2), 85-97. Bettman, J. R., Luce, M. F., & Payne, J. W. (1998). Constructive Consumer Choice Processes. Journal of Consumer Research, 25(3), 187-217. Bitter, S., & Grabner-Kräuter, S. (2016). Consequences of customer engagement behavior: when negative Facebook posts have positive effects. Electronic Markets, 26(3), 219-231. Blanton, H. (2001). Evaluating the self in the context of another: the three-selves model of social comparison assimilation and contrast. In G. B. Moskowitz (Ed.), Cognitive social psychology: The Princeton symposium on the legacy and future of social cognition (pp. 75-87). Mahwah: R Erlbaum. Bleier, A., & Eisenbeiss, M. (2015). Personalized Advertising Effectiveness: The Interplay of What, When, and Where. Marketing Science, 34(5), 669-688. Bodoff, D., & Ho, S. Y. (2015). Effectiveness of Website Personalization: Does the Presence of Personalized Recommendation Cannibalize Sampling of Other Items? International Journal of Electronic Commerce, 20(2), 208-235.

Boerman, S. C., & Kruikemeier, S. (2016). Consumer responses to promoted tweets sent by brands and political parties. Computers in Human Behavior, 65, 285-294. Boerman, S. C., Kruikemeier, S., & Zuiderveen Borgesius, F. J. (2017). Online Behavioral Advertising: A Literature Review and Research Agenda. Journal of Advertising, 46(3), 363-376. Bol, N., Kruikemeier, S., Boerman, S. C., Strycharz, J., Dienlin, T., Sax, M., & Helberger, N. (2018). Understanding the Effects of Personalization as a Privacy Calculus: Analyzing Self-Disclosure

211 References

Across Health, News and Commerce Contexts. Journal of Computer-Mediated Communication, 23(1), 370-388. boyd, d. m., & Ellison, N. B. (2007). Social Network Sites: Definition, History, and Scholarship. Journal of Computer-Mediated Communication, 13(1), 210-230. doi:10.1111/j.1083-6101.2007.00393.x Bright, L. F., & Daugherty, T. (2012). Does customization impact advertising effectiveness? An exploratory study of consumer perceptions in advertising in customized online environments. Journal of Marketing Communications, 18(1), 19-37. Brinson, N. H., & Eastin, M. S. (2016). Juxtaposing the persuasion knowledge model and privacy paradox: An experimental look at advertising personalization, public policy and public understanding. Cyberpsychology: Journal of Psychosocial Research on Cyberspace, 10(1). Brinson, N. H., Eastin, M. S., & Cicchirillo, V. J. (2018). Reactance to Personalization: Understanding the Drivers Behind the Growth of Ad Blocking. Journal of Interactive Advertising, 18(2), 136-147. Brown, J., Broderick, A. J., & Lee, N. (2007). Word of Mouth Communication within Online Communities: Conceptualizing the Online Social Network. Journal of Interactive Marketing, 21(3), 2-20. Brown, J. J., & Reingen, P. H. (1987). Social Ties and Word-of-Mouth Referral Behavior. Journal of Consumer Research, 14(3), 350-362. Brüggen, E. C., Foubert, B., & Gremler, D. D. (2011). Extreme Makeover: Short- and Long-Term Effects of a Remodeled Servicescape. Journal of Marketing, 75(5), 71-87. Bruner, G. C. I., & Kumar, A. (2000). Web commercials and advertising hierarchy-of-effects. Journal of Advertising Research, 40(1/2), 35-42. Bucher, T. (2017). The algorithmic imaginary: exploring the ordinary affects of Facebook algorithms. Information, Communication & Society, 20(1), 30-44. Buda, R., & Zhang, Y. (2000). Consumer product evaluation: the interactive effect of message framing, presentation order, and source credibility. Journal of Product & Brand Management, 9(4), 229- 242. Bušljeta Banks, I., Dens, N., & De Pelsmacker, P. (2016). The influence of gender on the effectiveness of probability markers in advertising. International Journal of Advertising, 35(4), 682-705. Buzeta, C., De Pelsmacker, P., & Dens, N. (2018). Motivations to use different social media types and their impact on consumer-brand engagement. Paper presented at the 17th International Conference on Research in Advertising, Valencia, Spain. Cacioppo, J. T., Petty, R. E., Feinstein, J. A., & Jarvis, W. B. G. (1996). Dispositional Differences in Cognitive Motivation: The Life and Times of Individuals Varying in Need for Cognition. Psychological Bulletin, 119(2), 197-253. Cadwalladr, C., & Graham-Harrison, E. (2018). Revealed: 50 millioin Facebook profiles harvested for Cambridge Analytica in major data breach. The Guardian. Retrieved from https://www.theguardian.com/uk-news/2019/mar/17/cambridge-analytica-year-on-lesson-in- institutional-failure-christopher-wylie Čaić, M., Mahr, D., Aguirre, E., de Ruyter, K., & Wetzels, M. (2015). "Too Close for Comfort": The Negative Effects of Location-Based Advertising. In I. Bušljeta Banks, P. De Pelsmacker, & S. Okazaki (Eds.), Advances in Advertising Research (Vol.V): Extending the Boundaries of Advertising (pp. 103-111). Wiesbaden: Springer Gabler. Cauberghe, V., Geuens, M., & De Pelsmacker, P. (2011). Context effects on TV programme-induced interactivity and telepresence on advertising responses. International Journal of Advertising, 30(4), 641-663. Chang, S. E., Liu, A. Y., & Shen, W. C. (2017). User trust in social networking services: A comparison of Facebook and LinkedIn. Computers in Human Behavior, 69, 207-217. Chang, Y.-T., Yu, H., & Lu, H.-P. (2015). Persuasive messages, poplarity cohesion, and market diffusion in social media marketing. Journal of Business Research, 68(4), 777-782.

212 References

Chellappa, R. K., & Sin, R. G. (2005). Personalization versus Privacy: An Empirical Examination of the Online Consumer's Dilemma. Information Technology and Management, 6(2-3), 181-202. Chen, H.-T. (2018). Revisiting the privacy paradox on social media with an extended privacy calculus model: The effect of privacy concerns, privacy self-efficacy, and social capital on privacy management. American behavioral scientist, 62(10), 1392-1412. Chen, Q., Clifford, S. J., & Wells, W. D. (2002). Attitude Toward the Site II: New Information. Journal of Advertising Research, 42(2), 33-45. Chen, Q., & Rodgers, S. (2006). Development of an Instrument to Measure Web Site Personality. Journal of Interactive Advertising, 7(1), 4-46. Chen, Q., & Wells, W. D. (1999). Attitude Toward the Site. Journal of Advertising Research, 39(5), 27- 37. Chen, T.-W., & Sundar, S. S. (2018). "This App Would Like to Use Your Current Location to Better Serve You": Importance of User Assent and System Transparency in Personalized Mobile Services. Paper presented at the CHI Conference on Human Factors in Computing Systems. Cheung, C. M. K., Lee, M. K. O., & Rabjohn, N. (2008). The impact of electronic word-of-mouth. Internet Research, 18(3), 229-247. Cheung, C. M. K., & Thadani, D. R. (2012). The impact of electronic word-of-mouth communication: A literature analysis and integrative model. Decision Support Systems, 54(1), 461-470. doi:http://dx.doi.org/10.1016/j.dss.2012.06.008 Cheung, M. Y., Luo, C., Sia, C. L., & Chen, H. (2009). Credibility of Electronic word-of-Mouth: Informational and Normative Determinants of On-Line Consumer Recommendations. International Journal of Electronic Commerce, 13(4), 9-38. Chiu, H.-C., Hsieh, Y.-C., Kao, Y.-H., & Lee, M. (2007). The Determinants of Email Receivers' Disseminating Behaviors on the Internet. Journal of Advertising Research, 47(4), 524-534. Cho, C.-H. (1999). How Advertising Works on the WWW: Modified Elaboration Likelihood Model. Journal of Current Issues and Research in Advertising, 21(1), 33-50. Chu, S.-C. (2011). Viral Advertising in Social Media: Participation in Facebook Groups and Responses among College-Aged Users. Journal of Interactive Advertising, 12(1), 30-43. Chu, S.-C., & Kim, J. (2018). The current state of knowledge on electronic word-of-mouth in advertising research. International Journal of Advertising, 37(1), 1-13. Chu, S.-C., & Kim, Y. (2011). Determinants of consumer engagement in electronic word-of-mouth (eWOM) in social networking sites. International Journal of Advertising, 30(1), 47-75. Chun, J. W., & Lee, M. J. (2016). Increasing individuals' involvement and WOM intention on Social Networking Sites: Content matters! Computers in Human Behavior, 60, 223-232. Codol, J.-P., Jarymowicz, M., Kaminska-Feldman, M., & Szuster-Zbrojewicz, A. (1989). Asymmetry in R the estimation of interpersonal distance and identity affirmation. European Journal of Social Psychology, 19(1), 11-22. Cooler Insights. (2016, 28/09/2016). How to Optimize Your Digital Marketing Funnel. Retrieved from https://coolerinsights.com/2016/09/how-to-optimize-your-digital-marketing-funnel/ Culnan, M. J., & Armstrong, P. K. (1999). Information Privacy Concerns, Procedural Fairness, and Impersonal Trust: An Empirical Investigation. Organization Science, 10(1), 104-115.

Dance, G. J. X., Confessore, N., & LaForgia, M. (2018). Facebook Gave Device Makers Deep Access to Data on Users and Friends. Retrieved from https://www.nytimes.com/interactive/2018/06/03/technology/facebook-device-partners-users- friends-data.html De Bruyn, A., & Lilien, G. L. (2008). A multi-stage model of word-of-mouth influence through viral marketing. International Journal of Research in Marketing, 25(3), 151-163.

213 References

De Cannière, M. H., De Pelsmacker, P., & Geuens, M. (2009). Relationship Quality and the Theory of Planned Behavior models of behavioral intentions and purchase behavior. Journal of Business Research, 62(1), 82-92. De Keyzer, F., Dens, N., & De Pelsmacker, P. (2015). Is this for me? How Consumers Respond to Personalized Advertising on Social Network Sites. Journal of Interactive Advertising, 15(2), 124- 134. De Keyzer, F., Dens, N., & De Pelsmacker, P. (2017). Don't be so emotional! How tone of voice and service type affect the relationship between message valence and consumer responses to WOM in social media. Online Information Review, 41(7), 905-920. De Keyzer, F., Dens, N., & De Pelsmacker, P. (2018). Personalized advertising on Facebook: the role of perceived relevance, intrusiveness, information control and privacy protection. Paper presented at the 16th International Conference on Research in Advertising (ICORIA), Valencia, Spain. De Keyzer, F., Dens, N., & De Pelsmacker, P. (2019). Qualitative Study on Personalized Advertising on Social Networking Sites. University of Antwerp. De Keyzer, F., Kruikemeier, S., & van Noort, G. (2019). Going Too Far? Consumer Responses Towards Personalized Content on Facebook From Different Sources. Paper presented at the 69th Annual ICA Conference, Washington, D.C., USA. De Meulenaer, S., Dens, N., & De Pelsmacker, P. (2015). Which cues cause consumers to perceive brands as more global? A conjoint analysis. International Marketing Review, 32(6), 606-626. De Pelsmacker, P., Geuens, M., & Anckaert, P. (2002). Media Context and Advertising Effectiveness: The Role of Context Appreciation and Context/Ad Similiarity. Journal of Advertising, 31(2), 49- 61. De Pelsmacker, P., Geuens, M., & Van Den Bergh, J. (2007). Marketing Communications: A European Perspective: Pearson Education. De Veirman, M., Cauberghe, V., & Hudders, L. (2017). Marketing through Instagram influencers: the impact of numbers of followers and product divergence on brand attitude. International Journal of Advertising, 36(5), 798-828. de Vries, L., Gensler, S., & Leeflang, P. S. H. (2012). Popularity of brand posts on brand fan pages: An investigation of the effects of social media marketing. Journal of Interactive Marketing, 26(2), 83-91. Demmers, J., van Dolen, W. M., & Weltevreden, J. W. J. (2018). Handling Consumer Messages on Social Networking Sites: Customer Service or Privacy Infringement? International Journal of Electronic Commerce, 22(1), 8-35. Dens, N., De Pelsmacker, P., & Purnawirawan, N. (2015). "We(b)care": How review set balance moderates the appropriate response strategy to negative online reviews. Journal of Service Management, 26(3), 486-515. Dhar, R., & Wertenbroch, K. (2000). Consumer Choice Between Hedonic and Utilitarian Goods. Journal of Marketing Research, 37(1), 60-71. Dienlin, T., & Metzger, M. J. (2016). An Extended Privacy Calculus Model for SNSs: Analyzing Self- Disclosure and Self-Withdrawal in a Representative U.S. Sample. Journal of Computer-Mediated Communication, 21(5), 368-383. Dijkstra, A. (2005). Working mechanisms of computer-tailored health education: Evidence from smoking cessation. Health Education Research, 20(5), 527-539. Dinev, T., & Hart, P. (2006). An Extended Privacy Calculus Model for E-Commerce Transactions. Information Systems Research, 17(1), 61-80. Dodds, W. B., Monroe, K. B., & Grewal, D. (1991). Effects of price, brand, and store information on buyers' product evaluations. Journal of Marketing Research, 28(3), 307-319. Doh, S.-J., & Hwang, J.-S. (2009). How Consumers Evaluate eWOM (Electronic Word-of-Mouth) Messages. Cyberpsychology & Behavior, 12(2), 193-197.

214 References

Duncan, T., & Moriarty, S. E. (1998). A Communication-Based Marketing Model for Managing Relationships. Journal of Marketing, 62(2), 1-13. Eberly, M. B., Holley, E. C., Johnson, M. D., & Mitchell, T. R. (2011). Beyond Internal and External: A Dyadic Theory of Relational Attributions. Academy of Management Review, 36(4), 731-753. Edwards, S. M., Li, H., & Lee, J.-H. (2013). Forced Exposure and Psychological Reactance: Antecedents and Conseques of the Perceived Intrusiveness of Pop-Up Ads. Journal of Advertising, 31(3), 83- 95. Eisingerich, A. B., Chun, H. H., Liu, Y., Jia, H., & Bell, S. J. (2015). Why recommend a brand face-to- face but not on Facebook? How word-of-mouth on online social sites differs from traditional word-of-mouth. Journal of Consumer Psychology, 25(1), 120-128. doi:10.1016/j.jcps.2014.05.004 Ellison, N. B., Steinfield, C., & Lampe, C. (2007). The Benefits of Facebook "Friends:" Social Capital and College Students' Use of Online Social Network Sites. Journal of Computer-Mediated Communication, 12(4), 1143-1168. Escalas, J. E. (2004). Narrative Processing: Building Consumer Connections to Brands. Journal of Consumer Psychology, 14(1&2), 168-180. Escalas, J. E., & Bettman, J. R. (2003). You Are What They Eat: The Influence of Reference Groups on Consumers' Connections to Brands. Journal of Consumer Psychology, 13(3), 339-348. Euromonitor International. (2015, July). Fragrances in Belgium. Retrieved from http://www.euromonitor.com/fragrances-in-belgium/report European Commitee. (2018). EU General Data Protection Regulation. Retrieved from https://ec.europa.eu/info/law/law-topic/data- protection/reform_nl?pk_source=google_ads&pk_medium=paid&pk_campaign=gdpr2019 Facebook. (2008). Ad Audiences. Retrieved from https://www.facebook.com/help/433385333434831?locale=en_US Facebook. (2014). Facebook Reports Third Quarter Results 2014. Retrieved from http://investor.fb.com/releasedetail.cfm?ReleaseID=878726 Facebook. (2018). Company Info - Stats. Retrieved from http://newsroom.fb.com/company-info/ Facebook Inc. (2019a). Ad Targeting. Retrieved from https://www.facebook.com/business/ads/ad- targeting Facebook Inc. (2019b). Company Info - Stats. Retrieved from https://newsroom.fb.com/company- info/ Facebook Inc. (2019c). Facebook Reports Fourth Quarter and Full Year 2018 Results. Retrieved from Menlo Park, California: https://s21.q4cdn.com/399680738/files/doc_financials/2018/Q4/Q4- 2018-Earnings-Release.pdf R Facebook Inc. (2019d). Facebook Reports Second Quarter 2019 Results. Retrieved from Menlo Park, California: https://s21.q4cdn.com/399680738/files/doc_financials/2019/Q2/FB-Q2-2019- Earnings-Release.pdf Fang, Y.-H. (2014). Beyond the Credibility of Electronic Word of Mouth: Exploring eWOM Adoption on Social Networking Sites from Affective and Curiosity Perspectives. International Journal of Electronic Commerce, 18(3), 67-102. Feick, L., & Higie, R. A. (1992). The effects of preference heterogeneity and source characteristics on ad processing and judgements about endorsers. Journal of Advertising, 21(2), 9-24. Felbermayr, A., & Nanopoulos, A. (2016). The Role of Emotions for the Perceived Usefulness in Online Customer Reviews. Journal of Interactive Marketing, 36, 60-75. Feldman, J. M., & Lynch, J. G. J. (1988). Self-Generated Validity and Other Effects of Measurement on Belief, Attitude, Intention, and Behavior. Journal of Applied Psychology, 73(3), 421-435. Festinger, L. (1954). A theory of social comparison processes. Human Relations, 7(2), 117-140.

215 References

Floyd, K., Freling, R., Alhoqail, S., Cho, H. Y., & Freling, T. (2014). How Online Product Reviews Affect Retail Sales: A Meta-Analysis. Journal of Retailing, 90(2), 217-232. Fogg, B. (2003). Prominence-Interpretation Theory: Explaining How People Assess Credibility Online. Paper presented at the CHI'03; Extended Abstracts on Human Factors in Computing Systems, 722-723. Fogués, R. L., Such, J. M., Espinosa, A., & Garcia-Fornes, A. (2014). BFF: A tool for eliciting tie strength and user communities in social networking services. Information Systems Frontiers, 16(2), 225- 237. Fornell, C., & Larcker, D. F. (1981). Evaluating structural equation models with unobservable variables and measurement error. Journal of Marketing Research, 18(1), 39-50. Fournier, S. (1998). Consumers and Their Brands: Developing Relationship Theory in Consumer Research. Journal of Consumer Research, 24(4), 343-373. doi:Doi 10.1086/209515 Franke, N., Keinz, P., & Steger, C. J. (2009). Testing the Value of Customization: When Do Customers Really Prefer Products Tailored to Their Preferences? Journal of Marketing, 73(5), 103-121. Franke, N., & Schreier, M. (2008). Product Uniqueness as a Driver of Customer Utility in Mass Customization. Marketing Letters, 19(2), 93-107. Franke, N., Schreier, M., & Kaiser, U. (2010). The "I Designed It Myself" Effect in Mass Customization. Management Science, 56(1), 125-140. Friestad, M., & Wright, P. (1994). The Persuasion Knowledge Model: How People Cope with Persuasion Attempts. Journal of Consumer Research, 21(1), 1-31. Geuens, M., & De Pelsmacker, P. (2017). Planning and Conducting Experimental Advertising Research and Questionnaire Design. Journal of Advertising, 46(1), 83-100. Gilbert, E., & Karahalios, K. (2009). Predicting Tie Strength With Social Media. Paper presented at the SIGCHI conference on human factors in computing systems, ACM. Gilly, M. C., Graham, J. L., Wolfinbarger, M. F., & Yale, L. J. (1998). A Dyadic Study of Interpersonal Information Search. Journal of the Academy of Marketing Science, 1998(26), 2. Gironda, J. t., & Korgaonkar, P. K. (2018). iSpy? Tailored versus Invasive Ads and Consumers' Perceptions of Personalized Advertising. Electronic Commerce Research and Applications, 29, 64-77. Goldsmith, R. E., & Lafferty, B. A. (2002). Consumer response to Web sites and their influence on advertising effectiveness. Internet Research: Electronic Networking Applications and Policy, 12(4), 313-328. Gordon, M. E., McKeage, K., & Fox, M. A. (1998). Relationship Marketing Effectiveness: The Role of Involvement. Psychology & Marketing, 15(5), 443-359. Grabner-Kräuter, S., & Waiguny, M. K. (2015). "I Believe More in Factual Reviews" - But not so Much when the Reviewer Is Similar to the Reader and the Product is Hedonic. In I. Bušljeta Banks, P. De Pelsmacker, & S. Okazaki (Eds.), Advances in Advertising Research (Vol. 5, pp. 89-102). Wiesbaden: Springer Fachmedien. Granovetter, M. S. (1973). The strength of weak ties. American Journal of Sociology, 78(6), 1360-1380. Gvili, Y., & Levy, S. (2016). Antecedents of attitudes toward eWOM communication: differences across channels. Internet Research, 26(5), 1030-2243. Hair, J. F., Black, W. C., Babin, B. J., Anderson, R. e., & Tatham, R. L. (2010). Multivariate data analysis: a global perspective (7 ed.). Upper Saddle River, New Jersey: Prentice Hall. Hair, J. F. J., Black, W. C., Bain, B. J., & Anderson, R. E. (2014). Multivariate Data Analysis (Vol. 7th edition). Essex, United Kingdom: Pearson Eduction Limited. Hajli, N. (2016). Ethical Environment in the Online Communities by Information Credibility: A Social Media Perspective. Journal of Business Ethics, 1-12.

216 References

Ham, C.-D. (2016). Exploring how consumers cope with online behavioral advertising. International Journal of Advertising, 36(4), 632-658. Hanneman, R. A., & Riddle, M. (2005). Introduction to social network methods Retrieved from http://www.faculty.ucr.edu/~hanneman/nettext/ Haugtvedt, C. P., Petty, R. E., & Cacioppo, J. T. (1992). Need for Cognition and Advertising: Understanding the Role of Personality Variables in Consumer Behavior. Journal of Consumer Psychology, 1(3), 239-260. Hawkins, R. P., Kreuter, M., Resnicow, K., Fishbein, M., & Dijkstra, A. (2008). Understanding Tailoring in Communication about Health. Health Education Research, 23(3), 454-466. Hayes, A. F. (2013). Introduction to Mediation, Moderation, and Conditional Process Analysis: A Regression-Based Approach (Vol. 1). New York: The Guildford Press. Hayes, A. F. (2018). Introduction to Mediation, Moderation and Conditional Process Analysis (Second Edition ed.). New York: The Guilford Press. Helgeson, J. G., & Supphellen, M. (2004). A Conceptual and Measurement Comparison of Self-Congruity and Brand Personality: The Impact of Socially Desirable Responding. International Journal of Market Research, 46(2), 205-233. Hellén, K., & Sääksjärvi, M. (2011). Happiness as a Predictor of Service Quality and Commitment for Utilitarian and Hedonic Services. Psychology & Marketing, 28(9), 934-957. Hennig-Thurau, T., Gwinner, K. P., Walsh, G., & Gremler, D. D. (2004). Electronic Word-of-Mouth Via Consumer-Opinion Platforms: What Motivates Consumer to Articulate Themselves on the Internet? Journal of Interactive Marketing, 18(1), 38-52. Herr, P. M., Kardes, F. R., & Kim, J. (1991). Effects of Word-of-Mouth and Product-Attribute Information on Persuasion: An Accessibility-Diagnosticity Perspective. Journal of Consumer Research, 17(4), 454-462. Higgins, S. F., Mulvenna, M. D., Bond, R. B., McCartan, A., Gallagher, S., & Quinn, D. (2018). Multivariate Testing Confirms the Effect of Age-Gender Congruence on Click-Through Rates from Online Social Network Digital Advertisements. Cyberpsychology, Behavior and Social Networking, 21(10), 646-654. High on Life. (2019). Ad Details. Retrieved from https://www.facebook.com/ads/library/?active_status=all&ad_type=all&country=BE&q=High %20On%20Life&view_all_page_id=340411786304159 Hirsch, J. B., Kang, S. K., & Bodenhausen, G. V. (2012). Personalized Persuasion: Tailoring Persuasive Appeals to Recipients' Personality Traits. Psychological Science, 23(6), 578-581. Hirschman, E. C., & Holbrook, M. B. (1982). Hedonic Consumption: Emerging Concepts, Methods and Propositions. Journal of Marketing, 46(3), 92-101. R Hollebeek, L. D., Glynn, M. S., & Brodie, R. J. (2014). Consumer Brand Engagement in Social Media: Conceptualization, Scale Development and Validation. Journal of Interactive Marketing, 28(2), 149-165. Howard, D. J., & Kerin, R. A. (2004). The Effects of Personalized Product Recommendations on Advertising Response Rates: The „Try This. It Works!" Technique. Journal of Consumer Psychology, 14(3), 271-279. Hoy, M. G., & Milne, G. (2010). Gender Differences in Privacy-Related Measures for Young Adult Facebook Users. Journal of Interactive Advertising, 10(2), 28-45. Hsieh, Y.-C., Chiu, H.-C., & Chiang, M.-Y. (2005). Maintaining a committed online customer: A study across search-experience-credence products. Journal of Retailing, 81(1), 75-82. Hsu, T. (2018, 21/03/2018). For Many Facebook Users, a 'Last Straw' That Led Them To Quit. The New York Times. Retrieved from https://www.nytimes.com/2018/03/21/technology/users-abandon- facebook.html

217 References

Hu, L.-t., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6(1), 1-55. Huang, L.-S., Chou, Y.-J., & Lan, I.-T. (2007). Effects of Perceived Risk, Message Types, and Reading Motives on the Acceptance and Transmission of Electronic Word-of-Mouth Communication. Contemporary Management Research, 3(4), 299-312. IHS Markit. (2015). The Economic Value of Behavioural Targeting in Digital Advertising. Retrieved from https://datadrivenadvertising.eu/wp- content/uploads/2017/09/BehaviouralTargeting_FINAL.pdf Instagram. (2019). Build Your Business on Instagram. Retrieved from https://business.instagram.com/advertising/ Interactive Advertising Bureau. (2015). IAB Display Advertising Guidelines. Interbrand. (2017). Best Global Brands 2017 - Rankings. Retrieved from http://interbrand.com/best- brands/best-global-brands/2017/ranking/ Iyer, G., Soberman, D., & Villas-Boas, J. M. (2005). The Targeting of Advertising. Marketing Science, 24(3), 461-476. Jiang, Y., & Wang, C. L. (2006). The impact of affect on service quality and satisfaction: the moderation of service contexts. Journal of Services Marketing, 20(4), 211-218. Kalyanaraman, S., & Sundar, S. S. (2006). The Psychological Appeal of Personalized Content in Web Portals: Does Customization Affect Attitudes and Behavior? Journal of Communication, 56(1), 110-132. Kang, C., & Frenkel, S. (2018, 4/04/2018). Facebook Says Cambridge Analytica Harvested Data of Up to 87 Million Users. The New York Times. Retrieved from https://www.nytimes.com/2018/04/04/technology/mark-zuckerberg-testify-congress.html Kaplan, A. M., & Haenlein, M. (2010). Users of the world, unite! The challenges and opportunities of Social Media. Business Horizons, 53(1), 59-68. Karakaya, F., & Barners, N. G. (2010). Impact of online reviews of customer care experience on brand or company selection. Journal of Consumer Marketing 27(5), 447-457. Karjaluoto, H., Munnukka, J., & Kiuru, K. (2016). Brand love and positive word of mouth: the moderating effects of experience and price. Journal of Product & Brand Management, 25(6), 527-537. Karmakar, I., & Webster, J. (2018). Personalization of an Environmental Message: Developing a Measure. Paper presented at the Thirty-Ninth International Conference on Information Systems, San Francisco. Kaspar, K., Weber, S. L., & Wilbers, A.-K. (2019). Personally relevant online advertisements: Effects of demographic targeting on visual attention and brand evaluation. PLoS ONE, 14(2), 1-18. Katz, E., & Lazarsfeld, P. F. (1955). Personal influence: the part played by people in the flow of mass communication. Glencoe, IL: Free Press. Kees, J., Berry, C., Burton, S., & Sheehan, K. (2017). An Analysis of Data Quality: Professional Panels, Student Subject Pools, and Amazon's Mechanical Turk. Journal of Advertising, 46(1), 141-155. Kelley, H. H. (1973). The Processes of Causal Attribution. American Psychologist, 28(2), 107-128. Kelly, L., Kerr, G., & Drennan, J. (2010). Avoidance of Advertising in Social Networking Sites: The Teenage Perspective. Journal of Interactive Advertising, 10(2), 16-27. Kelly, L., Kerr, G., & Drennan, J. (2017). Privacy concerns on social networking sites: a longitudinal study. Journal of Marketing Management, 33(17-18), 1465-1489. Kemp, S. (2019). Digital 2019. Retrieved from https://datareportal.com/reports/digital-2019-q2-global- digital-statshot

218 References

Ketelaar, P., Bernritter, S. F., van't Riet, J., Hühn, A. E., van Woudenberg, T. J., Müller, B. C. N., & Janssen, L. (2017). Disentangling location-based advertising: the effects of location congruency and medium type on consumers' ad attention and brand choice. International Journal of Advertising, 36(2), 356-367. Ketelaar, P. E., Bernritter, S. F., Van Woudenberg, T. J., Rozendaal, E., Konig, R. P., Hühn, A. E., . . . Janssen, L. (2018). "Opening" location-based mobile ads: How openness and location congruency of location-based ads weaken negative effects of intrusiveness on brand choice. Journal of Business Research, 91, 277-285. Kim, D., Jang, S. S., & Adler, H. (2015). What drives café customers to spread eWOM?: Examining self- relevant value, quality value, and opinion leadership. International Journal of Contemporary Hospitality Management, 27(2), 261-282. Kim, E., & Kang, H. (2014). Brand followers' retweeting behavior on Twitter: How brand relationships influence brand electronic word-of-mouth. Computers in Human Behavior, 37, 18-25. Kim, H., & Huh, J. (2017). Perceived Relevance and Privacy Concern Regarding Online Behavior Advertising (OBA) and Their Role in Consumer Responses. Journal of Current Issues & Research in Advertising, 38(1), 92-105. Kim, J., & Gambino, A. (2016). Do we trust the crowd or information system? Effects of personalization and bandwagon cues on users' attitudes and behavioral intentions toward a restaurant recommendation website. Computers in Human Behavior, 65, 369-379. Kim, J., & Gupta, P. (2012). Emotional expressions in online user reviews: How they influence consumer's product evaluations. Journal of Business Research, 65(7), 985-992. Kim, N. Y., & Sundar, S. (2012). Personal Relevance Versus Contextual Relevance: The Role of Relevant Ads in Personalized Websites. Journal of Media Psychology, 24(3), 89-101. Kim, N. Y., & Sundar, S. S. (2008). What's Relevance Got to Do With it? A Moderated Mediation Exploration of the Appeal of Personalization in Websites. Paper presented at the Annual Meeting International Communication Association, Montreal. Kim, S., Kandampully, J., & Bilgihan, a. (2018). The influence of eWOM communications: An appliciation of online social network framework. Computers in Human Behavior, 80, 243-254. Kim, S., Lee, J., & Yoon, D. (2015). Norms in Social Media: The Application of Theory of Reasoned Action and Personal Norms in Predicting Interactions With Facebook Page Like Ads. Communication Research Reports, 32(4), 322-331. Kim, Y., Park, Y., Lee, Y., & Park, K. (2018). Do we always adopt Facebook friends' eWOM postings? The role of social identity and threat. International Journal of Advertising, 37(1), 86-104. Kim, Y. J., & Han, J. (2014). Why smartphone advertising attracts customers: A model of Web advertising, flow, and personalization. Computers in Human Behavior, 33, 256-269. King, M. F., & Balasubramanian, S. K. (1994). The Effects of Expertise, End Goal, and Product Type on R Adoption of Preference Formation Strategy. Journal of the Academy of Marketing Science, 22(2), 146-159.

King, R. A., Racherla, P., & Bush, V. D. (2014). What We Know and Don't Know About Online Word-of- Mouth: A Review and Synthesis of the Literature. Journal of Interactive Marketing, 28(3), 167- 183. Klein, K., & Melnyk, V. (2016). Speaking to the mind or the heart: effects of matching hedonic versus utilitarian arguments and products. Marketing Letters, 27(1), 131-142. Klein, L. R. (1998). Evaluating the Potential of Interactive Media through a New Lens: Search versus Experience Goods. Journal of Business Research, 41(3), 195-203. Koo, W., Cho, E., & Kim, Y.-K. (2014). Actual and ideal self-congruity affecting consumers' emotional and behavioral responses toward an online store. Computers in Human Behavior, 36, 147-153. Kotler, P., & Keller, K. L. (2006). Identifying Market Segments and Targets Marketing Management (12th ed., pp. 239-271). Upper Saddle River, New Jersey: Pearson Education Inc.

219 References

Kramer, T. (2007). The Effect of Measurement Task Transparency on Preference Construction and Evaluations of Personalized Recommendations. Journal of Marketing Research, 44(2), 224-233. Kramer, T., Spolter-Weisfeld, S., & Thakkar, M. (2007). The Effect of Cultural Orientation on Consumer Response to Personalization. Marketing Science, 26(2), 246-258. Kruikemeier, S., van Noort, G., Vliegenthart, R., & de Vreese, C. H. (2013). Getting closer: The effects of personalized and interactive political communication. European Journal of Communication, 28(1), 53-66. Ku, Y.-C., Chu, T.-H., & Tseng, C.-H. (2013). Gratifications for Using CMC Technologies: A Comparison among SNS, IM, E-mail. Computers in Human Behavior, 29(1), 226-234. Kumar, V., & Gupta, S. (2016). Conceptualizing the Evolution and Future of Advertising. Journal of Advertising, 45(3), 302-317. Kusumasondjaja, S., Shanka, T., & Marchegiani, C. (2012). Credibility of online reviews and initial trust: the roles of reviewer's identity and review valence. Journal of Vacation Marketing, 18(3), 185- 195. Kwon, E. S., & Sung, Y. (2011). Follow Me! Global Marketers' Twitter Use. Journal of Interactive Advertising, 12(1), 4-16. Laczniak, R. N., DeCarlo, T. E., & Ramaswami, S. N. (2001). Consumers' Responses to Negative Word- of-Mouth Communication: An Attribution Theory Perspective. Journal of Consumer Psychology, 11(1), 57-73. Lapowsky, I. (2018). The 21 (And Counting) Biggest Facebook Scandals of 2018. Retrieved from https://www.wired.com/story/facebook-scandals-2018/ Lasswell, H. (1948). The structure and function of communication in society. In L. Bryson (Ed.), The communiction of idea (pp. 37-51). New York: Harper and Row. Laufer, R. S., & Wolfe, M. (1977). Privacy as a Concept and a Social Issue: A Multidimensional Developmental Theory. Journal of Social Issues, 33(3), 22-42. Lazarsfeld, P., & Merton, R. K. (1954). Friendship as a social process: a substantive and methodological analysis. In M. Berger, T. Abel, & C. H. Page (Eds.), Freedom and Control in Modern Society (pp. 18-66). New York: Van Nostrand. Lee, J., Park, D.-H., & Han, I. (2008). The effect of negative online consumer reviews on product attitude: An information processing view. Electronic Commerce Research and Applications, 7(3), 341-352. Lee, K.-T., & Koo, D.-M. (2012). Effects of attribute and valence of e-WOM on message adoption: Moderating roles of subjective knowledge and regulatory focus. Computers in Human Behavior, 28(5), 1974-1984. Lee, M., & Youn, S. (2009). Electronic word of mouth (eWOM): How eWOM platforms influence consumer product judgement. International Journal of Advertising, 28(3), 473-499. Lee, S., Kim, K. J., & Sundar, S. S. (2015). Customization in location-based advertising: Effects of tailoring source, locational congruity, and product involvement on ad attitudes. Computers in Human Behavior, 51, 336-343. Lee, Y.-J., & Ahn, H.-Y. (2013). Interaction Effects of Perceived Sponsor Motives and Facebook Credibility on Willingness to Visit Social Cause Facebook Page. Journal of Interactive Advertising, 13(1), 41-52. Lee, Y. H. (2000). Manipulating Ad Message Involvement through Information Expectancy: Effects on Attitude Evaluation and Confidence. Journal of Advertising, 29(2), 29-43. Li, C. (2016). When does web-based personalization really work? The distinction between actual personalization and perceived personalization. Computers in Human Behavior, 54, 25-33. doi:10.1016/j.chb.2015.07.049

220 References

Li, C., & Liu, J. (2017). Effects of using social networking sites in different languages: Does Spanish or English make a difference? Computers in Human Behavior, 74, 257-264. doi:10.1016/j.chb.2017.04.031 Li, H., Edwards, S. M., & Lee, J.-H. (2002). Measuring the Intrusiveness of Advertisements: Scale Development and Validation. Journal of Advertising, 31(2), 37-47. Li, H., Sarathy, R., & Xu, H. (2011). The role of affect and cognition on online consumers' decision to disclose personal information to unfamiliar online vendors. Decision Support Systems, 51(3). Li, M., Huang, L., Tan, C.-H., & Wei, K.-K. (2013). Helpfulness of Online Product Reviews as Seen By Consumers: Source and Content Features. International Journal of Electronic Commerce, 17(4), 101-136. Li, T., & Unger, T. (2012). Willing to pay for quality personalization? Trade-off between quality and privacy. European Journal of Information Systems, 21(6), 621-642. Lin, Y.-H., Hsu, C.-L., Chen, M.-F., & Fang, C.-H. (2017). New gratifications for social word-of-mouth spread via mobile SNSs: Uses and gratifications approach with a perspective of media technology. Telematics and Informatics, 34(4), 382-397. Liu-Thompkins, Y. (2012). Seeding Viral Content: The Role of Message and Network Factors. Journal of Advertising Research, 52(2), 59-72. Liu, S. Q., & Mattila, A. S. (2017). Airbnb: Online targeted advertising, sense of power, and consumer decisions. International Journal of Hospitality Management, 60, 33-41. Liu, S. S., & Stout, P. A. (1987). Effects of Message Modality and Appeal on Advertising Acceptance. Psychology & Marketing, 4(3), 167-187. Liu, Y. (2006). Word of Mouth for Movies: Its Dynamics and Impact on Box Office Revenue. Journal of Marketing, 70(3), 74-89. López-López, I., & Parra, J. F. (2016). Is a most helpful eWOM review really helpful? The impact of conflicting aggregate valence and consumer's goals on product attitude. Internet Research, 26(4), 827-844. Magna Global. (2013, Dec 9, 2013). Magna Global Advertising Forecasts: 2014. Retrieved from http://news.ipgmediabrands.com/ipgmediabrands/press-releases/magna-global-advertising- forecast-2014-ipg-mediabrands.print Malheiros, M. (2013). User Behavior in Personal Data Disclosure. Retrieved from http://discovery.ucl.ac.uk/1427369/2/MIGUEL_MALHEIROS_844263_phd_thesis_malheiros.pd f._COPYRIGHT_REMOVED..pdf Malheiros, M., Jennett, C., Patel, S., Brostoff, S., & Sasse, M. A. (2012). Too Close for Comfort: A Study on the Effectiveness and Acceptability of Rich-Media Personalized Advertising. Paper presented at the SIGCHI conference on human factors in computing systems. R Malhotra, N. H., Kim, S. S., & Agarwal, J. (2004). Internet Users' Information Privacy Concerns (IUIPC): The Construct, the Scale, and a Causal Model. Information Systems Research, 15(4), 336-355. Marchand, A., Hennig-Thurau, & Wiertz, C. (2017). Not all digital word of mouth is created equal: Understanding the respective impact of consumer reviews and microblogs on new product success. International Journal of Research in Marketing, 34(4), 336-354. Marketo. (2018). The AI-Empowered Marketer: How Artificial Intelligence Enhances the Marketer. Retrieved from https://www.marketo.com/infographics/the-ai-empowered-marketer-how- artificial-intelligence-enhances-the-marketer/ Marotta, V., Zhang, K., & Acquisti, A. (2015). The Welfare and Allocative Impact of Targeted Advertising. Paper presented at the Thirty Sixth International Conference on Information Systems, Fort Worth. Marsden, P. V., & Campbell, K. E. (1984). Tie Strength. Social Forces, 63(2), 482-501.

221 References

Maslowska, E., Smit, E., & van den Putte, B. (2011, June 23rd - 25th 2011). Examining the (In)Effectivenessof Personalized Communication. Paper presented at the 10th International Conference on Research in Advertising, Berlin. Maslowska, E., Smit, E. G., & van den Putte, B. (2016). It Is All in the Name: A Study of Consumers' Responses to Personalized Communication. Journal of Interactive Advertising, 16(1), 74-85. doi:10.1080/15252019.2016.1161568 Maslowska, E., van den Putte, B., & Smit, E. G. (2011). The Effectiveness of Personalized E-mail Newsletters and the Role of Personal Characteristics. Cyberpsychology, Behavior and Social Networking, 14(12), 765-770. Matz, S. C., Kosinski, M., Nave, G., & Stillwell, D. J. (2017). Psychological targeting as an effective approach to digital mass persuasion. Paper presented at the National Academy of Sciences. Maurer, C., & Wiegmann, R. (2011). Effectiveness of Advertising on Social Network Sites: A Case Study on Facebook. In R. Law, M. Fuchs, & F. Ricci (Eds.), Information and Communication Techologies in Tourism 2011 (pp. 485-498). Innsbruck, Austria: Springer Verlag-Wien. McAndrew, F. T., & Koehnke, S. S. (2016). On the nature of creepiness. New Ideas in Psychology, 43, 10-15. McCarthy, E. J. (1960). Basic Marketing: A Managerial Approach. Homewood, Illinois: Richard D. Irwan Inc., . McCroskey, J. C., Hamilton, P. R., & Weiner, A. N. (1974). The Effect of Interaction Behavior On Source Credibility, Homophily and Interpersonal Attraction. Human Communication Research, 1(1), 42- 52. McCroskey, L. L., McCroskey, J. C., & Richmond, V. P. (2006). Analysis and Improvement of the Measurement of Interpersonal Attraction and Homophily. Communication Quarterly, 54(1), 1- 31. McLuhan, M. (1964). Understanding Media: The Extensions of Man. New York: McGraw-Hill. McMurtry, J. (2017). Taking Personal Relevance Beyond the Message. Retrieved from https://www.targetmarketingmag.com/post/taking-personal-relevance-beyond-the-message/ Mcpherson, M., Smith-lovin, L., & Cook, J. M. (2001). Birds of a Feather: Homophily in Social Networks. Annual Review of Sociology, 27, 415-444. Menon, G., Raghubit, P., & Schwarz, N. (1995). Behavioral Frequency Judgments: An Accessibility- Diagnosticity Framework. Journal of Consumer Research, 22(2), 212-228. Metzger, M. J. (2007). Making sense of credibility on the Web: Models for evaluating online information and recommendations for future research. Journal of the American Society for Information Science and Technology, 58(13), 2078-2091. Meyers-Levy, J., & Loken, B. (2015). Revisiting gender differences: What we know and what lies ahead? Journal of Consumer Psychology, 25(1), 129-149. Meyers-Levy, J., & Sternthal, B. (1991). Gender Differences in the Use of Message Cues and Judgments. Journal of Marketing Research, 28(1), 84-96. Miron, A. M., & Brehm, J. W. (2006). Reactance Theory - 40 Years Later. Zeitschrift für Sozialpsychologie, 37(1), 9-18. Moe, W. W., & Schweidel, D. A. (2014). Social Media Intelligence. New York, USA: Cambridge University Press. Moeller, J., Trilling, D., Helberger, N., Irion, K., & de Vreese, C. (2016). Shrinking core? Exploring the differential agenda setting power of traditional and personalized news media. Info, 18(6), 26- 41. Money, R. B., Gilly, M. C., & Graham, J. L. (1998). Explorations of National Culture and Word-of-Mouth Referral Behavior in the Purchase of Industrial Services in the United States and Japan. Journal of Marketing, 62(4), 76-87.

222 References

Montgomery, K. C. (2015). Youth and surveillance in the Facebook era: Policy interventions and social implications. Telecommunications Policy, 39, 711-786. Moore, E., & Bowman, G. (2006). Of Friends and Family: How Do Peers Affect the Development of Intergenerational Influences? NA - Advances in Consumer Research (Vol. 33, pp. 536-542). Duluth, MN: Association for Consumer Research. Moran, G., & Muzellec, L. (2014). eWOM credibility on social networking sites: A framework. Journal of Marketing Communications, 1-13. doi:10.1080/13527266.2014.969756 Morimoto, M., & Chang, S. (2006). Consumers' Attitudes Toward Unsolicited Commercial E-mail and Postal Direct Mail Marketing Methods. Journal of Interactive Advertising, 7(1), 1-11. Mpinganjira, M., & Maduku, D. K. (2019). Ethics of mobile behavioral advertising: Antecedents and outcomes of perceived ethical value of advertised brands. Journal of Business Research, 95, 464-478. doi:https://doi.org/10.1016/j.jbusres.2018.07.037 Mudambi, S. M., & Schuff, D. (2010). What makes a helpful online review? A study of customer reviews on amazon.com. MIS Quarterly, 34(1), 185-200. Muntiga, D. G., Moorman, M., & Smit, E. G. (2011). Introducing COBRAs. International Journal of Advertising, 30(1), 13-46. Nelson, P. (1970). Information and Consumer Behavior. Journal of Political Economy, 78(2), 311-329. Nielsen. (2014). E-commerce: Evolution or Revolution in the Fast-Moving Consumer Goods World? Retrieved from Nisbett, R. E., & Wilson, T. D. (1977). Telling More Than We Can Know: Verbal Reports on Mental Processes. Psychological Review, 84(3), 231-259. doi:Doi 10.1037/0033-295x.84.3.231 Nissenbaum, H. (2010). Privacy in Context: Technology, Policy and the Integrity of Social Life. Stanford, CA: Stanford University Press. Noar, S. M., Benac, C. N., & Harris, M. S. (2007). Does Tailoring Matter? Meta-Analytic Review of Tailored Print Health Behavior Change Interventions. Psychological Bulletin, 133(4), 673-693. Noar, S. M., Harrington, N. G., & Aldrich, R. S. (2009). The Role of Message Tailoring in the Development of Persuasive Health Communication Messages. In C. S. Beck (Ed.), Communication Yearbook 33 (pp. 73-133). New York: Routledge. Ohanian, R. (1990). Construction and Validation of a Scale to Measure Celebrity Endorsers' Perceived Expertise, Trustworthiness, and Attractiveness. Journal of Advertising, 19(3), 39-52. Okazaki, S., & Taylor, C. R. (2013). Social media and international advertising: theoretical challenges and future directions. International Marketing Review, 30(1), 56-71. Ozcelik, A. B., & Varnali, K. (2019). Effectiveness of online behavioral targeting: A psychological perspective. Electronic Commerce Research and Applications, 33, 100819. R Palazon, M., Delgado-Ballester, E., & Sicilia, M. (2018). Fostering brand love in Facebook brand pages. Online Information Review, 43(5), 710-727. Pariser, E. (2012). The Filter Bubble: What the Internet Is Hiding from You. New York, NY: Penguin Press. Park, C., & Lee, T. M. (2009). Information direction, website reputation and eWOM effect: A moderating role of product type. Journal of Business Research, 62(1), 61-67. Park, D.-H., & Lee, J. (2008). eWOM overload and its effect on consumer behavioral intention depending on consumer involvement. Electronic Commerce Research and Applications, 7(4), 386-398. Park, H., Rodgers, S., & Stemmle, J. (2011). Health Organizations' Use of Facebook for Health Advertising and Promotion. Journal of Interactive Advertising, 12(1), 62-77. Park, S.-B., & Park, D.-H. (2013). The Effect of Low- versus High-Variance in Product Reviews on Product Evaluation. Psychology & Marketing, 30(7), 543-554.

223 References

Pavlou, P. A., & Stewart, D. W. (2000). Measuring the Effects and Effectiveness of Interactive Advertising: A Research Agenda. Journal of Interactive Advertising, 1(1), 62-78. Pentina, I., Ainsworth, A. B., & Zhang, L. (2018). Exploring effect of source similarity, message valence, and reviewer regulatory focus on yelp review persuasiveness and purchase intentions. Journal of Marketing Communications, 24(2), 125-145. doi:10.1080/13527266.2015.1005115 Pentina, I., Gammoh, B. S., Zhang, L., & Mallin, M. (2013). Drivers and Outcomes of Brand Relationship Quality in the Context of Online Social Networks. International Journal of Electronic Commerce, 17(3), 63-86. Percy, L., & Rossiter, J. R. (1992). A Model of Brand Awareness and Brand Attitude Advertising Strategies. Psychology & Marketing, 9(4), 263-274. Petty, R. E., Barden, J. S., & Wheeler, C. S. (2002). The Elaboration Likelihood Model: Developing Health Promotions for Sustained Behavioral Change. In R. J. DiClemente, R. A. Crosby, & M. Kegler (Eds.), Emerging theories in health promotion practice and research. San Francisco, CA: Jossey- Bass. Petty, R. E., & Briñol, P. (2010). Attitude change. In R. F. Baumeister & E. J. Finkel (Eds.), Advanced social psychology: The state of the science (pp. 217-259). Oxford, UK: Oxford University Press. Petty, R. E., & Cacioppo, J. T. (1984). The Effects of Involvement on Responses to Argument Quantity and Quality: Central and Peripheral Routes to Persuasion. Journal of Personality and Social Psychology, 46(1), 69-81. Petty, R. E., & Cacioppo, J. T. (1986). The Elaboration Likelihood Model of Persuasion. In L. Berkowitz (Ed.), Advances in Experimental Social Psychology (Vol. 19, pp. 123-205): Academic Press. Pfiffelmann, J., Dens, N., & Soulez, S. (2019). Personalized advertisements with integration of names and photograph: An eye-tracking experiment. Journal of Business Research, In press. doi:https://doi.org/10.1016/j.jbusres.2019.08.017. Pinterest. (2019). Advertise on Pinterest. Retrieved from https://help.pinterest.com/en/business/topics/advertise-on-pinterest Podsakoff, P. M., MacKenzie, S. B., Lee, J.-Y., & Podsakoff, N. P. (2003). Common Method Biases in Behavioral Research: A Critical Review of the Literature and Recommended Remedies. Journal of Applied Psychology, 88(5), 879-903. Pornpitakpan, C. (2004). The Persuasiveness of Source Credibility: A Critical Review of Five Decades' Evidence. Journal of Applied Social Psychology, 34(2), 243-281. Purnawirawan, N., De Pelsmacker, P., & Dens, N. (2012). Balance and sequence in online reviews: How perceived usefulness affects attitudes and intentions. Journal of Interactive Marketing, 26(4), 244-255. Purnawirawan, N., Dens, N., & De Pelsmacker, P. (2014). Expert reviewers beware! The effects of review set balance, review source and review content on consumer responses to online reviews. Journal of Electronic Commerce Research, 15(3), 162-178. Purnawirawan, N., Eisend, M., De Pelsmacker, P., & Dens, N. (2015). A Meta-Analytic Investigation of the Role of Valence in Online Reviews. Journal of Interactive Marketing, 31, 17-27. Quan-Haase, A., & Young, A. L. (2010). Uses and Gratifications of Social Media: A Comparison of Facebook and Instant Messaging. Bulletin of Science, Technology, and Society, 30(5), 350-361. Quaschning, S. (2013). The Role of Information in Consumers' Decisions: A Closer Look at Online Reviews and Product Lists. (Doctor in Applied Economics), Ghent University, Ghent, Belgium. Rajabi, M., Dens, N., & De Pelsmacker, P. (2015). Self-Brand Congruity and Brand Communication Marketing and Consumer Behavior: Concepts, Methodologies, Tools, and Applications (pp. 1411-1427): IGI Global. Reichelt, J., Sievert, J., & Jacob, F. (2014). How credibility affects eWOM reading: The influences of expertise, trustworthiness, and similarity on utilitarian an social functions. Journal of Marketing Communications, 20(1-2), 65-81.

224 References

Rimer, B. K., & Kreuter, M. W. (2006). Advancing Tailored Health Communication: A Persuasion and Message Effects Perspective. Journal of Communication, 56(1), 184-201. Rodgers, S., & Thorson, E. (2000). The Interactive Advertising Model: How Users Perceive and Process Online Ads. Journal of Interactive Advertising, 1(1), 42-61. Rogers, E. M. (1995). Diffusion of Innovations (4th ed.). New York: Free Press. Rogers, E. M., & Bhowmik, D. K. (1970). Homophily-Heterophily: Relational Concepts for Communication Research. The Public Opinion Quarterly, 34(4), 523-438. Romeo, J. B., & Debevec, K. (1992). An Investigation of Self-Referencing's Influence on Affective Evaluations. Advances in Consumer Research, 19, 290-295. Rossiter, J. R., & Percy, L. (1997). Advertising and promotion management (2 ed.). New York: McGraw- Hill. Ruef, M., Aldrich, H. E., & Carter, N. M. (2003). The Structure of Founding Teams: Homophily, Strong Ties and Isolation Among U.S. Entrepreneurs. American Sociological Review, 68(2), 195-222. Rui, H., Liu, Y., & Whinston, A. (2013). Whose and what chatter matters? The effect of tweets on movie sales. Decision Support Systems, 55(4), 863-870. Saxena, A., & Khanna, U. (2013). Advertising on Social Network Sites: A Structural Equation Modelling Approach. Vision, 17(1), 17-25. Schivinski, B., & Dabrowski, D. (2015). The impact of brand communication on brand equity through Facebook. Journal of Research in Interactive Marketing, 9(1), 31-53. Schultz, D. (2016). The Future of Advertising or Whatever We're Going to Call It. Journal of Advertising, 45(3), 276-285. Sen, S., & Lerman, D. (2007). Why are you telling me this? An examination into negative consumer reviews on the web. Journal of Interactive Marketing, 21(4), 76-94. Senecal, S., & Nantel, J. (2004). The influence of online product recommendations on consumers' online choices. Journal of Retailing, 80(2), 159-169. Shan, Y., & King, K. W. (2015). The Effect of Interpersonal Tie Strength and Subjective Norms on Consumers' Brand-Related eWOM Referral Intentions. Journal of Interactive Advertising, 15(1), 16-27. Shannon, C. E., & Weaver, W. (1949). The Mathematical Theory of Communication. Urbana: University of Illinois Press. Shavitt, S. (1990). The Role of Attitude Objects in Attitude Functions. Journal of Experimental Social Psychology, 26(March), 124-148. Shaw, A. (2018). Do People Really Look At Facebook Reviews? Here's Why It Matters. Retrieved from https://www.forbes.com/sites/forbescommunicationscouncil/2018/04/27/do-people-really- R look-at-facebook-reviews-heres-why-it-matters/#69f5e06e63b7 Sheeran, P. (2002). Intention - Behavior Relations: A Conceptual and Empirical REview. European Review of Social Psychology, 12(1), 1-36. Sheng, H., Nah, F. F.-H., & Siau, K. (2008). An Experimental Study on Ubiquitous Commerce Adoption: Impact of Personalization and Privacy Concerns. Journal of the Association for Information Systems, 9(6), 344-376. Sicilia, M., Ruiz, S., & Munuera, J. L. (2005). Effects of Interactivity in a Web Site: The Moderating Effect of Need for Cognition. Journal of Advertising, 34(3), 31-44. Simonson, I. (2005). Determinants of Customers' Responses to Customized Offers: Conceptual Framework and Research Propositions. Journal of Marketing, 69(1), 32-45. Simson, I., & Nowlis, S. M. (2000). The role of explanations and need for uniqueness in consumer decision making: unconventional choices based on reasons. Journal of Consumer Research, 27, 49-68.

225 References

Smit, E., van Noort, G., & Voorveld, H. A. M. (2014). Understanding online behavioural advertising: User knowledge, privacy concerns and online coping behaviour in Europe. Computers in Human Behavior, 32(March), 15-22. Sociable Labs. (2012). Social impact study: How consumer see it. Retrieved from https://www.slideshare.net/Briancrotty/sociable-labs-social-impact-consumer-study-3-252012 Sohl, S. J., & Moyer, A. (2007). Tailored interventions to promote mammography screening: a meta- analytic review. Preventive Medicine, 45(4), 252-261. Song, J. H., Kim, H. Y., Kim, S., Won Lee, S., & Lee, J.-H. (2016). Effects of personalized e-mail messages on privacy risk: Moderating roles of control and intimacy. Marketing Letters, 27(1), 89-101. Spears, N., & Singh, S. N. (2004). Measuring Attitude toward the Brand and Purchase Intentions. Journal of Current Issues & Research in Advertising, 26(2), 53-66. Srinivasan, S. S., Anderson, R., & Ponnavolu, K. (2002). Customer loyalty in e-commerce: an exploration of its antecedents and consequences. Journal of Retailing, 78(1), 41-50. Stadd, A. (2014). U.S. Social Advertising Spend Expected To Increase To $11 Billion By 2017. Retrieved from https://www.mediabistro.com/alltwitter/u-s-social-advertising_b53866 Statista, I. (2018). Most famous social network sites worlwide as of July 2018, ranked by active users (in millions). Retrieved from https://www.statista.com/statistics/272014/global-social- networks-ranked-by-number-of-users/ Statista, I. (2019). Most popular social network sites worlwide as of April 2019, ranked by active users (in millions). Retrieved from https://www.statista.com/statistics/272014/global-social- networks-ranked-by-number-of-users/ Statista Inc. (2016). Leading social media websites in the United States in October 2016, based on share of visits. Retrieved from https://www.statista.com/statistics/265773/market-share-of-the- most-popular-social-media-websites-in-the-us/ Statista Inc. (2019). Most popular social media websites in the United States in May 2019, based on share of visits. Retrieved from http://www.statista.com/statistics/265773/market-share-of- the-most-popular-social-media-websites-in-the-us/ Steffes, E. M., & Burgee, L. E. (2009). Social ties and online word of mouth. Internet Research, 19(1), 42-59. doi:10.1108/10662240910927812 Stevenson, J. S., Bruner, G. C. I., & Kumar, A. (2000). Webpage Background and Viewer Attitudes. Journal of Advertising Research, 40(1-2), 29-34. Strycharz, J., van Noort, G., Helberger, N., & Smith, E. G. (2019). Contrasting perspectives - practitioner's viewpoint on personalised marketing communication. European Journal of Marketing, 53(4), 635-660. Sundar, S. S., & Marathe, S. S. (2010). Personalization versus Customization: The Importance of Agency, Privacy, and Power Usage. Human Communication Research, 36(3), 298-322. Sweeney, J. C., & Brandon, C. (2006). Brand Personality: Exploring the Potential to Move from Factor Analytical to Circumplex Models. Psychology & Marketing, 23(8), 639-663. Sweeney, J. C., Soutar, G. N., & Mazzarol, T. (2008). Factors influencing word of mouth effectiveness: receiver perspectives. European Journal of Marketing, 42(3/4), 344-364. Sweeney, J. C., Soutar, G. N., & Mazzarol, T. (2012). Word of mouth: measuring the power of individual messages. European Journal of Marketing, 46(1/2), 237-225. Tam, K. Y., & Ho, S. Y. (2005). Web Personalization as a Persuasion Strategy: An Elaboration Likelihood Model Perspective. Information Systems Research, 16(3), 271-291. Tam, K. Y., & Ho, S. Y. (2006). Understanding the Impact of Web Personalization on User Information Processing and Decision Outcomes. MIS Quarterly, 30(4), 865-890.

226 References

Taylor, D. G., Lewin, J. E., & Strutton, D. (2011). Friends, Fans, and Followers: Do Ads Work on Social Networks. Journal of Advertising Research, 51(1), 258-275. Taylor, D. G., Strutton, D., & Thompson, K. (2012). Self-Enhancement as a Motivation for Sharing Online Advertising. Journal of Interactive Advertising, 12(2), 13-28. Teng, S., Wei, K., Wei, K., Goh, W., Yee, A., & Chong, L. (2014). Examining the antecedents of persuasive eWOM messages in social media. Online Information Review, 38(6), 746-768. Thelwall, M. (2009). Homophily in Myspace. Journal of the American Society for Information Science and Technology, 60(2), 219-231. Think With Google. (2017). Australia's Origin Energy Moves Ahead of the Pack With Life Events Targeting on YouTube. Retrieved from https://www.thinkwithgoogle.com/intl/en-aunz/advertising- channels/video/australias-origin-energy-moves-ahead-pack-life-events-targeting-youtube/ Thurman, N., Moeller, J., Helberger, N., & Trilling, D. (2018). My Friends, Editors, Algorithms, and I. Digital Journalism. Tran, T. P. (2017). Personalized ads on Facebook: An effective marketing tool for online marketers. Journal of Retailing and Consumer Services, 39, 230-242. doi:10.1016/j.jretconser.2017.06.010 Trusov, M., Bucklin, R. E., & Pauwels, K. (2009). Effects of Word-of-Mouth Versus Traditional Marketing: Findings from an Internet Social Networking Site. Journal of Marketing, 73(5), 90-102. Tucker, C. E. (2014). Social Networks, Personalized Advertising and Privacy Controls. Journal of Marketing Research, 51(5), 546-562. Tuna, T., Akbas, E., Aksoy, A., Canbaz, M. A., Karabiyik, U., Gonen, B., & Aygun, R. (2016). User characterization for online social networks. Social Network Analysis and Mining, 6(104), 1-28. Turow, J., King, J., Hoofnagle, C. J., Bleakley, A., & Hennessy, M. (2009). Americans Reject Tailored Advertising and Three Activities that Enable It. Retrieved from http://repository.upenn.edu/asc- papers/137 Twitter. (2019). Ad targeting. Retrieved from https://business.twitter.com/en/targeting.html Ur, B., Leon, P. G., Cranor, L. F., Shay, R., & Wang, Y. (2012). Smart, useful, scary, creepy: perceptions of online behavioral advertising. Paper presented at the proceedings of the eighth symposium on usable privacy and security. Uribe, R., Buzeta, C., & Velásquez, M. (2016). Sidedness, commercial intent and expertise in blog advertising. Journal of Business Research, 69(10), 4403-4410. Van den Broeck, E., Poels, K., & Walrave, M. (2015). Older and Wiser? Facebook Use, Privacy Concern, and Privacy Protection in the Life Stages of Emerging, Young and Middle Adulthood. social media + society, 1(2), 1-11. van Doorn, J., & Hoekstra, J. C. (2013). Customization of online advertising: the role of intrusiveness. Marketing Letters, 24(4), 339-351. R Van Ittersum, K. (2012). The effect of decision makers' time perspective on intention-behavior consistency. Marketing Letters, 23(1), 264-277. van Noort, G., & Willemsen, L. M. (2012). Online Damage Control: The Effects of Proactive Versus Reactive Webcare Interventions in Consumer-generated and Brand-generated Platforms. Journal of Interactive Marketing, 26(3), 131-140. doi:http://dx.doi.org/10.1016/j.intmar.2011.07.001

Verberckmoes, S., Poels, K., Dens, N., Herrewijn, L., & De Pelsmacker, P. (2016). When and why is perceived congruity important for in-game advertising in fantasy games? Computers in Human Behavior, 64(871-880). Vesanen, J. (2007). What is personalization? A conceptual framework. European Journal of Marketing, 41(5/6), 409-418. Voorveld, H. A. M. (2019). Brand Communication in Social Media: A Research Agenda. Journal of Advertising, 48(1), 14-26.

227 References

Voorveld, H. A. M., van Noort, G., Muntinga, D. G., & Bronner, F. (2018). Engagement with Social Media and Social Media Advertising: The Differentiating Role of Platform Type. Journal of Advertising, 47(1), 38-54. doi:10.1080/00913367.2017.1405754 Voss, K. E., Spangenberg, E. R., & Grohmann, B. (2003). Measuring the hedonic and utilitarian dimensions of consumer attitude. Journal of Marketing Research, 11, 310-320. Walrave, M., Poels, K., Antheunis, M. L., Van den Broeck, E., & van Noort, G. (2016). Like or dislike? Adolescents’ responses to personalized social network site advertising. Journal of Marketing Communications, 24(6), 599-616. doi:10.1080/13527266.2016.1182938 Wang, J.-C., & Chang, C.-H. (2013). How online social ties and product-related risks influence purchase intentions: A Facebook experiment. Electronic Commerce Research and Applications, 12(5), 337-346. Wang, S., Cunnigham, N. R., & Eastin, M. S. (2015). The Impact of eWOM Message Characteristics on the Perceived Effectiveness of Online Consumer Reviews. Journal of Interactive Advertising, 15(2), 151-159. Wang, T., Duong, T. D., & Chen, C. C. (2016). Intention to disclose personal information via mobile applications: A privacy calculus perspective. International Journal of Information Management, 36(4), 531-542. Wang, X., Yu, C., & Wei, Y. (2012). Social Media Peer Communication and Impacts on Purchase Intentions: A Consumer Socialization Framework. Journal of Interactive Marketing, 26(4), 198- 208. Wang, Z., Walther, J. B., Pingree, S., & Hawkins, R. P. (2008). Health Information, Credibility, Homophily, and Influence via the Internet: Web Sites Versus Discussion Groups. Health Information, 23(4), 358-368. Weathers, D., Sharma, S., & Wood, S. L. (2007). Effects of online communication practices on consumer perceptions of performance uncertainty for search and experience goods. Journal of Retailing, 83(4), 393-401. Webb, M. S., Hendricks, P. S., & Brandon, T. H. (2007). Expectancy priming of smoking cessation messages enhances the placebo effect of tailored interventions. Health Psychology, 26(5), 598- 609. Weiss, M. G., Ramakrishna, J., & Somma, D. (2006). Health-related stigma: rethinking concepts and interventions. Psychology, health & medicine, 11(3), 277-287. Wen, C., Tan, B. C. Y., & Chang, K., Ting-Ting. (2009). Advertising Effectiveness on Social Network Sites: An Investigation of Tie Strength, Endorser Expertise and Product Type on Consumer Purchase Intention. Paper presented at the International Conference on Information Systems. West, C. (2019). Social media demographics to drive your brand's online presence. Retrieved from https://sproutsocial.com/insights/new-social-media-demographics/#Facebook White, T. B., Zahay, D. L., Thorbjørnsen, H., & Shavitt, S. (2008). Getting Too Personal: Reactance to Highly Personalized Email Solicitation. Marketing Letters, 19(1), 39-50. Willemsen, L. M. (2013). Electronic word of mouth: Challenges for consumers and companies. (Ph.D.), University of Amsterdam, Amsterdam. Willemsen, L. M., Neijens, P. C., Bronner, F., & de Ridder, J. A. (2011). "Highly Recommended!" The Content Characteristics and Perceived Usefulness of Online Consumer Reviews. Journal of Computer-Mediated Communication, 17(1), 19-38. Windels, K., Heo, J., Jeong, Y., Porter, L., Jung, A. R., & Wang, R. (2018). My friend likes this brand: Do ads with social context attract more attention on social networking sites? Computers in Human Behavior, 84, 420-429. doi:10.1016/j.chb.2018.02.036 Wu, P. C. S., & Wang, Y.-C. (2011). The influences of electronic word-of-mouth message appeal and message source credibility on brand attitude. Asia Pacific Journal of Marketing and Logistics, 23(4), 448-472.

228 References

Xia, L., & Bechwati, N. N. (2008). Word of Mouse: the Role of Cognitive Personalization in Online Consumer Reviews. Journal of Interactive Advertising, 9(1), 3-13. Xie, W., & Karan, K. (2019). Consumers’ Privacy Concern and Privacy Protection on Social Network Sites in the Era of Big Data: Empirical Evidence from College Students. Journal of Interactive Advertising, online first. doi:https://doi.org/10.1080/15252019.2019.1651681 Xu, H., Oh, L.-B., & Teo, H.-H. (2009). Perceived effectiveness of text vs multimedia Location-Based Advertising messaging. International Journal of Mobile Communications, 7(2), 154-177. Yang, C. C., Tang, X., Dai, Q., Yang, H., & Jiang, L. (2013). Identifying Implicit and Explicit Relationships Through User Activities in Social Media. International Journal of Electronic Commerce, 18(2), 73-96. Young, S., & Kim, S. (2019). Understanding ad avoidance on Facebook: Antecedents and outcomes of psychological reactance. Computers in Human Behavior, 98, 232-244. Yu, J., & Cude, B. (2009). 'Hello, Mrs. Sarah Jones! We recommend this product!' Consumers' perceptions about personalized advertising: comparisons across advertisements delivered via three different types of media. International Journal of Consumer Studies, 33(4), 503-514. Zaichkowsky, J. L. (1985). Measuring the Involvement Construct. Journal of Consumer Research, 12(3), 341-352. Zarouali, B., Poels, K., Walrave, M., & Ponnet, K. (2018). The impact of regulatory focus on adolescents' evaluation of targeted advertising on social networking sites. International Journal of Advertising, 38(2), 316-335. Zeithaml, V. A. (1981). How Consumer Evaluation Processes Differ for Products and Services. In J. H. Donnelly & W. R. George (Eds.), Marketing of services (pp. 186-190). Chicago, IL: American Marketing Association. Zeng, F., Huang, L., & Dou, W. (2009). Social Factors in User Perceptions and Responses to Advertising in Online Social Networking Communities. Journal of Interactive Advertising, 10(1), 1-13. Zhang, B., & Xu, H. (2016). Privacy Nudges for Mobile Applications: Effects on the Creepiness Emotion and Privacy Attitudes. Paper presented at the 19th ACM Conference on Computer-Supported Cooperative Work & Social Computing, San Francisco, CA, USA. Zhao, X., Lynch, J. G. J., & Chen, Q. (2010). Reconsidering Baron and Kenny: Myths and Truths about Mediation Analysis. Journal of Consumer Research, 37(2), 197-206. Zhu, F., & Zhang, X. (2010). Impact of Online Consumer Reviews on Sales: The Moderating Role of Product and Consumer Characteristics. Journal of Marketing, 74(2), 133-148. Zhu, Y.-Q., & Chang, J.-H. (2016). The key role of relevance in personalized advertisment: Examining its impact on perceptions of privacy invasion, self-awareness, and continuous use intentions. Computers in Human Behavior, 65, 442-447. R Zou, P., Yu, B., & Hao, Y. (2011). Does the Valence of Online Consumer Reviews matter for Consumer Decision Making? The Moderating Role of Consumer Expertise? Journal of Computers, 6(3), 484-488. Zuiderveen Borgesius, F. J., Trilling, D. C., Moeller, J. E., Bodó, B., de Vreese, C. H., & Helberger, N. (2016). Should we worry about filter bubbles? Internet Policy Review, 5(1), 1-16. doi:10.14763/2016.1.401

229

Appendix

A Appendix

Appendix 1 – Stimulus Materials

1.A. Stimuli used in Chapter 2

Card 1 for a bank: Personalization elements used: location, life events, Facebook connections

232 Appendix

Card 2 for a bank: Personalization elements used: location, gender, age and Facebook connections

233 Appendix

Card 3 for a bank: Personalization elements used: age, life events, interests and Facebook connections

234 Appendix

Card 4 for a bank: Personalization elements used: location and interests and Facebook connections

235 Appendix

Card 5 for a bank: Personalization elements used: age

236 Appendix

Card 6 for a bank: Personalization elements used: location, age, gender, life events and interests

237 Appendix

Card 7 for a bank: Personalization elements used: gender, life events and interests

238 Appendix

Card 8 for a bank: Personalization elements used: gender and life events

239 Appendix

Card 1 for tableware: Personalization elements used: location, life events, Facebook connections

240 Appendix

Card 1 for a fashion retailer: Personalization elements used: location, life events, Facebook connections

241 Appendix

Card 1 for a restaurant: Personalization elements used: location, life events, Facebook connections

242 Appendix

Card 1 for a smartphone: Personalization elements used: location, life events, Facebook connections

243 Appendix

Card 1 for furniture: Personalization elements used: location, life events, Facebook connections

244

1.B. Stimuli used in Chapter 3

Condition 1: Non-personalized banner advertisement (Study 1)

Appendix

245

246 Condition 2: Personalized banner advertisement (Study 1) Appendix

Condition 1: Non-personalized banner advertisement (Study 2)

Appendix

247

248 Condition 2: Personalized banner advertisement (Study 2) Appendix

Appendix

1.C. Stimuli used in Chapter 4

Study 1 – Condition 1: Control condition

249 Appendix

Study 1 – Condition 2: Age-based personalization

250 Appendix

Study 1 – Condition 3: Gender-based personalization

251 Appendix

Study 1 – Condition 4: Interest-based personalization

252 Appendix

Study 2 - Condition 1: Non-personalized advertisement

253 Appendix

Study 2 - Condition 1: Control condition

254 Appendix

Study 2 - Condition 2: Advertisement personalized based on gender and age

255 Appendix

Study 2 - Condition 3: Advertisement personalized based on gender, age and interests

256 Appendix

Study 2 - Condition 4: Advertisement personalized based on gender, age, interests and life events

257 Appendix

Study 2 - Condition 5: Advertisement personalized based on gender, age, interests, life events and page-likes from friends

258

1.D. Stimuli used in Chapter 5

Low Medium High Health Imagine you would visit Facebook. There Imagine you would visit Facebook. There you Imagine you would visit Facebook. There you website you see an advertisement for a health see an advertisement for a health website see an advertisement for a health website website like gezondheidsplein.nl or like gezondheidsplein.nl or thuisarts.nl based gezondheidsplein.nl or thuisarts.nl based on thuisarts.nl based on your friends’ page- on your friends’ page-likes and your gender. your friends’ page-likes, your gender and likes. Your friends’ page-likes are used to Your friends’ page-likes and your gender are interests. Your friends’ page-likes, your show the advertisement. Before you used to show the advertisement. Before you gender and interests are used to show the continue with the questionnaire, we would continue with the questionnaire, we would advertisement. Before you continue with the like to ask you to imagine this like to ask you to imagine this questionnaire, we would like to ask you to advertisement. advertisement. imagine this advertisement.

Governmental Imagine you would visit Facebook. There Imagine you would visit Facebook. There you Imagine you would visit Facebook. There you website you see an advertisement for a see an advertisement for a governmental see an advertisement for a governmental governmental website, like website, like belastingdienst.nl or douane.nl website, belastingdienst.nl or douane.nl belastingdienst.nl or douane.nl based on based on your friends’ page-likes and your based on your friends’ page-likes, your your friends’ page-likes. Your friends’ gender. Your friends’ page-likes and your gender and interests. Your friends’ page- page-likes are used to show the gender are used to show the advertisement. likes, your gender and interests are used to advertisement. Before you continue with Before you continue with the questionnaire, show the advertisement. Before you continue the questionnaire, we would like to ask we would like to ask you to imagine this with the questionnaire, we would like to ask you to imagine this advertisement. advertisement. you to imagine this advertisement.

Online news Imagine you would visit Facebook. There Imagine you would visit Facebook. There you Imagine you would visit Facebook. There you website you see an advertisement for an online see an advertisement for an online see an advertisement for an online newspaper, like volkskrant.nl or nu.nl newspaper, like volkskrant.nl or nu.nl based newspaper, volkskrant.nl or nu.nl based on based on your friends’ page-likes. Your on your friends’ page-likes and your gender. your friends’ page-likes, your gender and friends’ page-likes are used to show the Your friends’ page-likes and your gender are interests. Your friends’ page-likes, your advertisement. Before you continue with used to show the advertisement. Before you gender and interests are used to show the the questionnaire, we would like to ask continue with the questionnaire, we would advertisement. Before you continue with the you to imagine this advertisement. like to ask you to imagine this questionnaire, we would like to ask you to advertisement. imagine this advertisement.

Appendix Online store Imagine you would visit Facebook. There Imagine you would visit Facebook. There you Imagine you would visit Facebook. There you you see an advertisement for an online see an advertisement for an online store, like see an advertisement for an online store, like 259 store, like zalando.nl or bol.com based on zalando.nl or bol.com based on your friends’ zalando.nl or bol.com based on your friends’

your friends’ page-likes. Your friends’ page-likes and your gender. Your friends’ page-likes, your gender and interests. Your

260 page-likes are used to show the page-likes and your gender are used to show friends’ page-likes, your gender and interests Appendix advertisement. Before you continue with the advertisement. Before you continue with are used to show the advertisement. Before

the questionnaire, we would like to ask the questionnaire, we would like to ask you you continue with the questionnaire, we

you to imagine this advertisement. to imagine this advertisement. would like to ask you to imagine this

advertisement.

Appendix

1.E. Stimuli used in Chapter 6

Condition 1: Positive, factual sWOM message about a hedonic service

261 Appendix

Condition 2: Negative, factual sWOM message about a hedonic service

262 Appendix

Condition 3: Positive, emotional sWOM message about a hedonic service

263 Appendix

Condition 4: Negative, emotional sWOM message about a hedonic service

264 Appendix

Condition 5: Positive, factual sWOM message about a utilitarian service

265 Appendix

Condition 6: Negative, factual sWOM message about a utilitarian service

266 Appendix

Condition 7: Positive, emotional sWOM message about a utilitarian service

267 Appendix

Condition 8: Negative, emotional sWOM message about a utilitarian service

268

1.F. Stimuli used in Chapter 7

Condition 1: Positive message from homophilious and credible source with a strong tie

Appendix

269

270 Condition 2: Positive message from heterophilious and credible source with a strong tie Appendix

Condition 3: Positive message from homophilious and non-credible source with a weak tie

Appendix

271

272 Condition 4: Positive message from heterophilious and non-credible source with weak tie Appendix

Condition 5: Positive message from homophilious and non-credible source with strong tie

Appendix

273

274 Condition 6: Positive message from heterophilious and non-credible source with strong tie Appendix

Condition 7: Positive message from homophilious and credible source with weak tie

Appendix

275

276 Condition 8: Positive message from heterophilious and credible source with weak tie Appendix

Condition 9: Negative message from homophilious and credible source with a strong tie

Appendix

277

278 Condition 10: Negative message from heterophilious and credible source with a strong tie Appendix

Condition 11: Negative message from homophilious and non-credible source with a weak tie

Appendix

280 Condition 12: Negative message from heterophilious and non-credible source with weak tie Appendix

Condition 13: Negative message from homophilious and credible source with strong tie

Appendix

281

282 Condition 14: Negative message from heterophilious and credible source with strong tie Appendix

Condition 15: Negative message from homophilious and non-credible source with weak tie

Appendix

283

284 Condition 16: Negative message from heterophilious and non-credible source with weak tie Appendix

Appendix

Appendix 2 – Questionnaires

2.A. Questionnaire used in Chapter 2 Dear Madam, dear Sir,

This study is part of a PhD research project conducted by the University of Antwerp, Belgium.

All data will be processed in full anonymity and will only be used in the context of academic research. There is no commercial purpose. Please note that the questionnaire contains images which may not fully be visible on a tablet or smartphone. Please make sure that you complete the survey on a laptop or computer.

If you have any questions or remarks concerning the questionnaire, please contact Freya De Keyzer via [email protected]

Thanks in advance for your cooperation.

Anne Pijnenburg

Freya De Keyzer prof.dr. Nathalie Dens prof.dr. Patrick De Pelsmacker

Which of the following social media do you use at least once a month?

o Facebook Appendix o Twitter o Instagram

o Snapchat o Whatsapp o YouTube

Are you …

o Male o Female A

What year were you born in?

…

What is the highest level of education you have completed? If you are a student, please indicate your current level of education.

Appendix

o Primary school o Middle (junior high) school o High school o Bachelors’ degree (Undergraduate) o Masters’ degree (Graduate) o PhD (doctorate) or postgraduate

Which state do you currently live in?

o [selection box with 50 U.S. states]

Please select the topic from the list below that you are most interested in.

o Socializing o Traveling o Sports o Golf o Art and Culture o Camping o Music o Hiking o Skiing o Reading o Gaming o Nature o Shopping o Going out

Which of the following life events has happened to you in the last 6 months?

o Got a new job o Started at a new school o Studied abroad o Started a new relationship o Got engaged o Got married o Celebrated an anniversary o Got a new child o Moved o Bought a home o Did home improvement o Traveled o None of the above

In the past week, on average, approximately how many days have you used Facebook?

o 1 day o 2 days o 3 days o 4 days o 5 days o 6 days o 7 days

286 Appendix

In the past week, on average, how many minutes per day have you spent on Facebook? o Less than 10 minutes o Between 11 and 30 minutes o Between 31 and 60 minutes o More than 1 hour, but less than 2 hours o More than 2 hours, but less than 3 hours o More than 3 hours

To what extent do you agree with the following statements?

Strongly Disagree Somewhat Neither Somewhat Agree Strongly disagree disagree disagree, agree agree nor agree Facebook is part of my o o o o o o o everyday activity. I feel out of touch when I o o o o o o o haven’t logged onto Facebook for a while. I would feel sorry if o o o o o o o Facebook shut down.

In the remainder of this research you will be presented with 8 different advertisements that you might encounter on Facebook. Please look at the advertisements attentively. The advertisements we ask you to evaluate will look highly similar, but they do differ (subtly). We therefore ask you to consider each ad carefully. For each advertisement you will be asked to rate this advertisement. Please rate each advertisement individually, as if you see it for the first time, disregarding your evaluations of any of the previous advertisements.

Appendix

Imagine that you would encounter the advertisement (suggested post) below in your Facebook

newsfeed, based on [personalization elements]

Eight personalized advertisements (randomized order)

Please rate the following statements. A Strongly Disagree Somewhat Neither Somewhat Agree Totally disagree disagree disagree, agree agree nor agree The ad is tailored to my o o o o o o o situation. I believe that this ad is o o o o o o o customized to my needs. This ad was targeted at me o o o o o o o as a unique individual.

Appendix

I believe that this ad is o o o o o o o customized to my characteristics. This ad was personalized o o o o o o o according to my profile. There was personal o o o o o o o information in the ad. The ad was targeted at me. o o o o o o o I could recognize myself in o o o o o o o the group the ad was targeted at.

Note this question was repeated for each of the eight advertisements.

[Product/service] to me is

-3 -2 -1 0 1 2 3 Unimportant o o o o o o o Important Meaningless o o o o o o o Meaningful Does not matter to o o o o o o o Does matter to me me

To what extent do you agree with the following statements?

Strongly Disagree Somewhat Neither Somewhat Agree Totally disagree disagree disagree, agree agree nor agree I can adequately evaluate [product/service] using information provided by the o o o o o o o retailer about its characteristics. I can adequately evaluate o o o o o o o the quality of [product/service] simply by reading information about [product/service]

To what extend to you agree with the following statements?

Strongly Disagree Somewhat Neither Somewhat Agree Totally disagree disagree disagree, agree agree nor agree It’s important for me to personally [try (on) o o o o o o o product/service] to evaluate it.

If I were to [choose/buy service/product], I would primarily do this because it is …

-3 -2 -1 0 1 2 3 Ineffective o o o o o o o Effective Not functional o o o o o o o Functional

288 Appendix

Impractical o o o o o o o Practical

If I were to [choose/buy service/product], I would primarily do this because it is …

-3 -2 -1 0 1 2 3 Not fun o o o o o o o Fun Dull o o o o o o o Exciting Unenjoyable o o o o o o o Enjoyable

Appendix

2.B. Questionnaire used in Chapter 3

Beste deelnemer,

In het kader van mijn Masterproef wil ik u vragen om een vragenlijst in te vullen. Het invullen zal maximum 10 minuten in beslag nemen. Gelieve steeds het bolletje te selecteren bij het antwoord dat het meest voor u van toepassing is. Verder gaan in de vragenlijst doet u door op de knop onderaan te klikken. Terug gaan in de vragenlijst is niet mogelijk.

Alvast hartelijk bedankt!

Bent u …

o Man o Vrouw

Hoe oud bent u?

… jaar

Wat is het hoogst behaalde diploma (voor studenten: wat is je huidige studie)?

o Lager secundair onderwijs o Hoger secundair onderwijs o Hogeschool o Universiteit o Andere …

Mogen wij u vragen om uw eigen naam en de naam van vijf van uw vrienden op te geven? Deze namen worden enkel gebruikt tijdens de survey en zullen achteraf niet worden bijgehouden. Indien u deze namen niet invult, zal u niet verder kunnen gaan met het invullen van de survey. Eigen naam: …. Vriend 1: … Vriend 2: … Vriend 3: … Vriend 4: … Vriend 5: …

Pagina met advertentie

Kan u aangeven in welke mate u akkoord gaat met de volgende stellingen?

290 Appendix

Helemaal Niet Enigszins Noch Enigszins Akkoord Helemaal niet akkoord niet niet akkoord niet akkoord akkoord akkoord, akkoord noch akkoord Het merk in de advertentie kan de claims o o o o o o o die het doet, waarmaken. Ik reageer gunstig o o o o o o o tegenover het merk in de advertentie. Ik heb positieve o o o o o o o gevoelens tegenover het merk in de advertentie. Ik houd niet van het o o o o o o o merk. (omgekeerd)

In welke mate was de informatie in de advertentie relevant voor u?

Irrelevant o o o o o o o Relevant

Was deze informatie nuttig voor u?

Niet nuttig o o o o o o o Nuttig

Was de informatie in de advertentie aangepast aan uw persoonlijk profiel?

Helemaal niet aangepast Volledig o o o o o o o

aangepast Appendix

Kan u aangeven in welke mate u op deze advertentie zou klikken?

Niet denkbaar o o o o o o o Denkbaar Onbestaande o o o o o o o Bestaand Onwaarschijnlijk o o o o o o o Waarschijnlijk Onmogelijk o o o o o o o Mogelijk Niet zeker o o o o o o o Zeer zeker Zeker niet klikken Zeker niet o o o o o o o klikken A

How do you feel about Facebook?

Helemaal Niet Enigszins Noch Enigszins Akkoord Helemaal niet akkoord niet niet akkoord niet akkoord akkoord akkoord, akkoord noch akkoord

Appendix

Ik ben trots als ik mensen o o o o o o o vertel dat ik Facebook gebruik. Ik zou het jammer vinden o o o o o o o als Facebook zou verdwijnen. Facebook maakt het o o o o o o o makkelijk voor mij om een relatie op te bouwen met mijn vrienden. Ik vind het gebruik van o o o o o o o Facebook een goede manier om mijn tijd te besteden.

292 Appendix

2.C. Questionnaires used in Chapter 4

Study 1 Beste,

In het kader van mijn masterproef nodig ik u graag uit om volgende enquête in te vullen. Ditz al slechts enkele minute in beslag nemen maar u zou er mij wel een groot plezier mee doen. Alle antwoorden zijn anoniem en worden voor geen andere doeleinden dan mijn masterproef gebruikt.

Alvast bedankt voor uw medewerking!

Wat is uw geslacht? o Vrouw o Man

Wat is uw leeftijd?

o 10-19 jaar o 20-29 jaar o 30-39 jaar o 40-49 jaar o 50-59 jaar o 60-69 jaar o 70 jaar of ouder

Wat is het hoogste diploma dat u heeft behaald? Appendix o Lager onderwijs o Middelbaar onderwijs o Professionele bachelor

o Academische bachelor o Master o Doctoraat o Andere …

Kies uit onderstaande lijst de interesse die het meest uw voorkeur draagt:

o Fietsen o Natuur A o Wandelen o Uitgaan o Skiën o Wellness o Shoppen o Cultuur

Appendix

Hoeveel tijd brengt u op een normale dag door op Facebook?

o Niet dagelijks o Minder dan 10 minuten o 11 and 30 minuten o 31 minuten en 1 uur o 1 - 2 uur o 2 - 3 uur o Meer dan 3 uur

U ziet hieronder een nieuwsoverzicht van een Facebook pagina. Beeld u in dat het uw eigen nieuwsoverzicht is. Bekijk de pagina en specifiek de advertentie aandachtig alvorens verder te gaan.

Pagina die nieuwsoverzicht bevat

Wat is uw mening over de advertentie?

Niet Enigszins niet Noch niet Enigszins Akkoord akkoord akkoord akkoord, akkoord noch akkoord Deze advertentie doet aanbevelingen o o o o o die passen bij mijn behoeften. Deze advertentie informeert me over o o o o o producten die op mijn maat gemaakt zijn. Deze advertentie is op maat van mijn o o o o o situatie gemaakt. Ik geloof dat deze advertentie is o o o o o aangepast aan mijn behoeften. Deze advertentie was gericht op mij o o o o o als uniek individu.

De informatie in de advertentie is …

-2 -1 0 1 2 Helemaal niet belangrijk o o o o o Heel belangrijk Helemaal niet relevant o o o o o Zeer relevant Helemaal niet significant o o o o o Zeer significant

Hoe komt de advertentie bij u over?

-2 -1 0 1 2 Helemaal niet afleidend o o o o o Zeer afleidend Helemaal niet Zeer verontrustend o o o o o verontrustend Helemaal niet geforceerd o o o o o Zeer geforceerd Helemaal niet storend o o o o o Zeer storend Helemaal niet opdringerig o o o o o Zeer opdringerig

In het algemeen, wat is uw gevoel over het reisagentschap Lovely Travel?

294 Appendix

-2 -1 0 1 2 Ik vind het helemaal niet Ik vind het heel leuk. o o o o o leuk. Het agentschap lijk Het agentschap lijkt o o o o o helemaal niet betrouwbaar. zeer betrouwbaar. Het gaat om een Het gaat om een agentschap van zeer lage o o o o o agentschap van zeer kwaliteit. hoge kwaliteit.

Appendix

Study 2

Dear Madam, dear Sir,

This study is part of a PhD research project conducted by the University of Antwerp, Belgium.

All data will be processed in full anonymity and will only be used in the context of academic research. There is no commercial purpose.

Please note that the questionnaire contains images which may not fully be visible on a tablet or smartphone. Please make sure that you complete the survey on a laptop or computer.

If you have any questions or remarks concerning the questionnaire, please contact Freya De Keyzer via [email protected]

Thanks in advance for your cooperation.

Anne Pijnenburg

Freya De Keyzer prof.dr. Nathalie Dens prof.dr. Patrick De Pelsmacker

Which of the following social media do you use at least once a month?

o Facebook o Twitter o Instagram o Snapchat o Whatsapp o YouTube

Are you …

o Male o Female

What year were you born in?

…

What is the highest level of education you have completed? If you are a student, please indicate your current level of education.

o Primary school o Middle (junior high) school

296 Appendix

o High school o Bachelors’ degree (Undergraduate) o Masters’ degree (Graduate) o PhD (doctorate) or postgraduate

Which state do you currently live in?

o [selection box with 50 U.S. states]

Please select the topic from the list below that you are most interested in.

o Socializing o Traveling o Sports o Golf o Art and Culture o Camping o Music o Hiking o Skiing o Reading o Gaming o Nature o Shopping o Going out

Which of the following life events has happened to you in the last 6 months?

o Got a new job o Started at a new school o Studied abroad o Started a new relationship Appendix o Got engaged o Got married

o Celebrated an anniversary o Got a new child o Moved o Bought a home o Did home improvement o Traveled o None of the above

In the past week, on average, approximately how many days have you used Facebook? A o 1 day o 2 days o 3 days o 4 days o 5 days o 6 days o 7 days

Appendix

In the past week, on average, how many minutes per day have you spent on Facebook?

o Less than 10 minutes o Between 11 and 30 minutes o Between 31 and 60 minutes o More than 1 hour, but less than 2 hours o More than 2 hours, but less than 3 hours o More than 3 hours

To what extent do you agree with the following statements?

On the next page you will see a Facebook news feed. Please imagine that you would see these messages in your own news feed. Please look at the page attentively. When you are done reading the newsfeed, please continue with the survey by clicking the “>>” button at the bottom of the page.

Page containing the personalized advertisement

Please describe your overall feeling about Radar (the advertised brand you just saw).

-3 -2 -1 0 1 2 3 Unappealing o o o o o o o Appealing Bad o o o o o o o Good Unpleasant o o o o o o o Pleasant Unfavorable o o o o o o o Favorable

To what extent do you agree with the following statements?

-3 -2 -1 0 1 2 3 I am likely to say I am likely to say negative things about positive things o o o o o o o Radar to other people. about Radar to other people. I am not likely to o o o o o o o I am likely recommend Radar to to a friend or a recommend colleague. Radar to a friend or a colleague.

To what extent do you agree with the following statements?

-3 -2 -1 0 1 2 3

298 Appendix

It is not likely that I It is likely that I o o o o o o o will click this ad. will click this ad.

Please indicate your opinion about Radar (the advertisers brand you just saw). Strongly Disagree Somewhat Neither Somewhat Agree Strongly disagree disagree disagree, agree agree nor agree Radar reflects who I am. o o o o o o o I can identify with Radar. o o o o o o o I feel a personal connection o o o o o o o with Radar. I could use Radar to o o o o o o o communicate who I am to other people. I think Radar would help o o o o o o o me become the type of person I want to be. I consider Radar to be “me” o o o o o o o (it reflects who I consider myself to be or the way I want to present myself to others). Radar suits me well o o o o o o o

Please rate the following statements. The information in the ad was … -3 -2 -1 0 1 2 3 Not important o o o o o o o Important Not relevant o o o o o o o Relevant Not significant o o o o o o o Significant

Please rate the following statements. The ad was …

Strongly Disagree Somewhat Neither Somewhat Agree Strongly Appendix disagree disagree disagree, agree agree nor agree Distracting o o o o o o o Forced o o o o o o o Interfering o o o o o o o Intrusive o o o o o o o Invasive o o o o o o o Obtrusive o o o o o o o Irritating o o o o o o o Annoying o o o o o o o

Please rate the following statements. The ad was … A

Strongly Disagree Somewhat Neither Somewhat Agree Strongly disagree disagree disagree, agree agree nor agree Enjoyable o o o o o o o Pleasant o o o o o o o Entertaining o o o o o o o

Appendix

Please rate the following statements.

Strongly Disagree Somewhat Neither Somewhat Agree Strongly disagree disagree disagree, agree agree nor agree I want to resist the o o o o o o o advertisement. I want to dismiss the o o o o o o o content of this advertisement. I want to avoid this kind of o o o o o o o advertisement.

Please rate the following statements.

Strongly Disagree Somewhat Neither Somewhat Agree Strongly disagree disagree disagree, agree agree nor agree The ad is tailored to my o o o o o o o situation. I believe this ad is o o o o o o o customized to my needs. This ad was targeted at me o o o o o o o as a unique individual. I believe that this ad is customized to my o o o o o o o characteristics. This ad was personalized o o o o o o o according to my profile.

Please rate the following statements.

Strongly Disagree Somewhat Neither Somewhat Agree Strongly disagree disagree disagree, agree agree nor agree Facebook has adequate o o o o o o o security features. I feel safe in my o o o o o o o transactions with Facebook. I feel like my privacy is o o o o o o o protected at Facebook. I trust Facebook will not misuse my personal o o o o o o o information. 2.D. Questionnaire used in Chapter 5

Beste deelnemer, U bent uitgenodigd deel te nemen aan een onderzoek dat wordt uitgevoerd onder verantwoordelijkheid van onderzoeksinstituut ASCoR, onderdeel van de Universiteit van Amsterdam. ASCoR doet wetenschappelijk onderzoek naar media en communicatie in de samenleving. Het onderzoek waarvoor wij uw medewerking hebben gevraagd, is getiteld ‘Onderzoek over reclame’. Aan dit onderzoek kunnen leden van het consumentenpanel van Panelclix deelnemen. Doel van het onderzoek is om inzicht te verkrijgen in de perceptie van reclame op Facebook.

300 Appendix

Tijdens dit onderzoek zullen wij u vragen een korte tekst te bekijken. Daarna zullen wij u vragen een vragenlijst in te vullen. Het onderzoek duurt circa 10 minuten en na afloop van het onderzoek ontvangt u een beloning zoals voorzien door Panelclix, als dank voor uw medewerking. Omdat dit onderzoek wordt uitgevoerd onder de verantwoordelijkheid van ASCoR, Universiteit van Amsterdam, heeft u de garantie dat: 1. Uw anonimiteit is gewaarborgd en dat uw antwoorden of gegevens onder geen enkele voorwaarde aan derden zullen worden verstrekt, tenzij u hiervoor van tevoren uitdrukkelijke toestemming hebt verleend. 2. U zonder opgaaf van redenen kunt weigeren mee te doen aan het onderzoek of uw deelname voortijdig kunt afbreken. Ook kunt u achteraf (binnen 24 uur na deelname) uw toestemming intrekken voor het gebruik van uw antwoorden of gegevens voor het onderzoek. 3. Deelname aan het onderzoek geen noemenswaardige risico’s of ongemakken voor u met zich meebrengt, en u niet met expliciet aanstootgevend materiaal zult worden geconfronteerd. 4. U uiterlijk 5 maanden na afloop van het onderzoek de beschikking over een onderzoeksrapportage kunt krijgen waarin de algemene resultaten van het onderzoek worden toegelicht. Voor meer informatie over dit onderzoek en de uitnodiging tot deelname kunt u te allen tijde contact opnemen met de projectleider Freya De Keyzer, ASCoR, Universiteit van Amsterdam, Postbus 15793, 1001 NG Amsterdam; [email protected]. Mochten er naar aanleiding van uw deelname aan dit onderzoek bij u toch klachten of opmerkingen zijn over het verloop van het onderzoek en de daarbij gevolgde procedure, dan kunt u contact opnemen met het lid van de Commissie Ethiek namens ASCoR, per adres: ASCoR secretariaat, Commissie Ethiek,

Universiteit van Amsterdam, Postbus 15793, 1001 NG Amsterdam; 020-525 3680; ascor-secr- [email protected]. Een vertrouwelijke behandeling van uw klacht of opmerking is daarbij gewaarborgd.

Wij hopen u hiermee voldoende te hebben geïnformeerd en danken u bij voorbaat hartelijk voor uw Appendix deelname aan dit onderzoek dat voor ons van grote waarde is. Met vriendelijke groet,

Freya De Keyzer, Sanne Kruikemeier en Guda van Noort

Amsterdam School of Communication Research, Universiteit van Amsterdam Hoeveel Facebookvrienden heeft u in totaal? (als u het niet precies weet, geef dan een schatting

…

A In de voorbije week hoeveel dagen heeft u Facebook gebruikt?

o 1 dag o 2 dagen o 3 dagen o 4 dagen o 5 dagen o 6 dagen o 7 dagen

Appendix

Op de volgende pagina krijgt u een verhaal te lezen waarin een situatie is beschreven.

Probeert u zich de situatie zo levendig mogelijk voor te stellen.

Wij vragen u om de situatie daarna te beoordelen.

Let op! U kunt niet terugkeren naar de pagina met de beschrijving van de situatie.

Op de volgende pagina kunt u pas na een paar seconden verder klikken.

Pagina met vignette

Wat is uw mening over de advertentie uit het verhaal? De advertentie is …

Niet belangrijk o o o o o o o Belangrijk Niet relevant o o o o o o o Relevant Betekenisloos o o o o o o o Betekenisvol

In hoeverre vind je de advertentie uit het verhaal …

Helemaal Heel niet erg 1 2 3 4 5 6 7 Griezelig o o o o o o o Storend o o o o o o o Alarmerend o o o o o o o

De [bron], zoals [voorbeeld 1] en [voorbeeld 2], uit het verhaal is …

Onaantrekkelijk o o o o o o o Aantrekkelijk Slecht o o o o o o o Goed Onaangenaam o o o o o o o Aangenaam Ongunstig o o o o o o o Gunstig

Bent u …

o Man o Vrouw

Wat is uw leeftijd (in jaren)?

…

Hartelijk dank voor uw deelname!

In dit onderzoek hebben we u een korte tekst laten zien. Deze tekst stelde telkens een verhaal voor waarin er bepaalde gegevens over u werden gebruikt in een advertentie op Facebook. Deze situatie was fictief. Het doel hiervan was te testen hoe mensen reageren op advertenties die gebruiken maken van persoonlijke gegevens.

302 Appendix

Aangezien dit onderzoek nog loopt, willen we u vragen het doel van dit onderzoek niet te delen met anderen.

Mocht u nog opmerkingen/vragen hebben naar aanleiding van dit onderzoek, dan kunt u deze hieronder geven/stellen.

Druk nogmaals op ‘>>’ zodat uw antwoorden worden verwerkt.

Appendix

2.E. Questionnaire used in Chapter 6

Dear Sir or Madam

This study is part of a research project conducted by the University of Antwerp, Belgium.

All data will be processed in full anonymity and will only be used in the context of academic research. There is no commercial purpose.

If you have any questions or remarks concerning the questionnaire, please contact Freya De Keyzer via [email protected]

Thanks in advance for your cooperation.

Freya De Keyzer

Prof.dr. Nathalie Dens

Prof.dr. Patrick De Pelsmacker

In the past week, on average, approximately how many days have you used Facebook

o 1 day o 2 days o 3 days o 4 days o 5 days o 6 days o 7 days

In the past week, on average, how many minutes per day have you spent on Facebook?

o Less than 10 minutes o Between 11 and 30 minutes o Between 31 and 60 minutes o More than 1 hour, but less than 2 hours o More than 2 hours, but less than 3 hours o More than 3 hours

To what extent do you agree with the following statements?

304 Appendix

To what extent do you think the post was negative or positive?

Very Negative Somewhat Neither Somewhat Positive Very negative negative negative nor positive positive positive o o o o o o o

How do you rate the content of the post by Michael Johnson -3 -2 -1 0 1 2 3 Intangible o o o o o o o Tangible Emotional o o o o o o o Rational Subjective o o o o o o o Objective Nonfactual o o o o o o o Factual Nonlogical o o o o o o o Logical

Please describe your overall feeling about Chromebar / Smartline.

-3 -2 -1 0 1 2 3 Unappealing o o o o o o o Appealing Bad o o o o o o o Good Unpleasant o o o o o o o Pleasant

Appendix Unfavorable o o o o o o o Favorable

To what extent do you agree with the following?

-3 -2 -1 0 1 2 3 I am likely to say I am likely to say positive negative things about things about o o o o o o o Chromebar/Smartline to Chromebar/Smartline to other people. other people. I am not likely to I am likely to recommend recommend o o o o o o o Chromebar/Smartline to Chromebar/Smartline to A a friend or colleague. a friend or colleague. I am likely to discourage I am likely to encourage friends and relatives to friends and relatives to o o o o o o o visit Chromebar/ visit Chromebar/ Smartline. Smartline.

A cell phone provider/bar to me is …

Appendix

-3 -2 -1 0 1 2 3 Unimportant o o o o o o o Important Meaningless o o o o o o o Meaningful Does not matter to me o o o o o o o Does not matter to me

If I were to use a cell phone provider / bar, I would primarily do this because it is …

-3 -2 -1 0 1 2 3 Ineffective o o o o o o o Effective Not functional o o o o o o o Functional Impractical o o o o o o o Practical

If I were to use a cell phone provider / bar, I would primarily do this because it is …

-3 -2 -1 0 1 2 3 Not fun o o o o o o o Fun Dull o o o o o o o Exciting Unenjoyable o o o o o o o Enjoyable

Are you

o Male o Female

What year were you born in?

…

What is the highest level of education you have completed? If you are a student, please indicate your current level of education.

o Primary school o Middle (junior high) school o High school o Bachelor’s degree (Undergraduate) o Masters’ degree (Graduate) o PhD (doctorate)

306 Appendix

2.F. Questionnaire used in Chapter 7

Dear Madam, dear Sir,

This study is part of a PhD research project conducted by the University of Antwerp, Belgium.

If you have any questions or remarks concerning the questionnaire, please contact Freya De Keyzer via [email protected]

Thanks in advance for your cooperation.

Freya De Keyzer

Prof.dr. Nathalie Dens

Prof.dr. Patrick De Pelsmacker

Are you

o Male o Female

Appendix What year were you born in?

…

What is the highest level of education you have completed? If you are a student, please indicate your current level of education. o Primary school o Middle (junior high) school o High school A o Bachelor’s degree (Undergraduate) o Masters’ degree (Graduate) o PhD (doctorate)

Which of the following ethnic groups do you belong to?

o Caucasian (non-hispanic white)

Appendix

o African American o Hispanic o Other

Please select the two topics from the list below that you are most interested in.

o Animals o Basketball o Cars o Classical music o Country music o Dancing o Food/cooking o Football o Going to the gym o Hiking o Massive Multiplayer Online Games (MMOG) o Mobile games o Poker o Pop music o Rock music o Running o Science o Skiing o Soccer o Traveling o Wine

In the past week, approximately how many days have you used Facebook?

o 0 days o 1 day o 2 days o 3 days o 4 days o 5 days o 6 days o 7 days

In the past week, on average, how many minutes per day have you spent on Facebook?

o Less than 10 minutes o Between 11 and 30 minutes o Between 31 and 60 minutes o More than 1 hours, but less than 2 hours o More than 2 hours, but less than 3 hours o More than 3 hours

To what extent do you agree with the following statements?

Strongly Disagree Somewhat Neither Somewhat Agree Strongly disagree disagree disagree, agree agree nor agree

308 Appendix

Facebook is part of my o o o o o o o everyday activity. I feel out of touch when I o o o o o o o haven’t logged onto Facebook for a while. I would feel sorry if o o o o o o o Facebook shut down.

To what extent do you agree with the following statements?

Strongly Disagree Somewhat Neither Somewhat Agree Strongly disagree disagree disagree, agree agree nor agree Facebook shares my o o o o o o o values. Facebook is like me. o o o o o o o Facebook has a lot in o o o o o o o common with me. My interests are similar to o o o o o o o the content I can find on Facebook

To what extent do you agree with the following statements?

Strongly Disagree Somewhat Neither Somewhat Agree Strongly disagree disagree disagree, agree agree nor agree I feel strongly linked to o o o o o o o Facebook. The relationship with o o o o o o o Facebook is important to

me. Appendix There are a lot of o o o o o o o activities/interaction between Facebook and me.

What is your opinion on Facebook with respect to movies? Facebook is …

-3 -2 -1 0 +1 +2 +3 Dishonest o o o o o o o Honest Unreliable o o o o o o o Reliable Untrustworthy o o o o o o o Trustworthy Not an expert o o o o o o o Expert A Inexperienced o o o o o o o Experienced Unknowledgeable o o o o o o o Knowledgeable Unqualified o o o o o o o Qualified Unskilled o o o o o o o Skilled

Instruction for low tie strength conditions

Appendix

On the next page you will see a Facebook profile. Please imagine that this is an actual Facebook profile of an acquaintance whom you rarely interact with. Please look at the profile attentively. Try to get a picture of what this person is like: what kind of things is this person interested in, what kind of people does this person interact with, etc.? When you are done, please continue with the survey by clicking the ">>" button at the bottom of the page. Instruction for high tie strength conditions

On the next page you will see a Facebook profile. Please imagine that this is an actual Facebook profile of a close friend whom you interact with almost daily. Please look at the profile attentively. Try to get a picture of what this person is like: what kind of things is this person interested in, what kind of people does this person interact with, etc.? When you are done, please continue with the survey by clicking the ">>" button at the bottom of the page.

Page containing the sWOM message

The movie mentioned by [name of the fictitious sWOM sender] was Dirty. Have you already seen the movie Dirty?

o Yes o No

To what extent do you think the post by [name of the fictitious sWOM sender] about the movie Dirty was negative or positive?

Very Negative Somewhat Neither Somewhat Positive Very negative negative negative nor positive positive positive o o o o o o o

To what extent do you agree with the following statements?

Strongly Disagree Somewhat Neither Somewhat Agree Strongly disagree disagree disagree, agree agree nor agree [name of the fictitious sWOM sender] shares my o o o o o o o values. [name of the fictitious o o o o o o o sWOM sender] is like me. [name of the fictitious o o o o o o o sWOM sender] has a lot in common with me.

310 Appendix

To what extent do you agree with the following statements?

Strongly Disagree Somewhat Neither Somewhat Agree Strongly disagree disagree disagree, agree agree nor agree I want my relationship with [name of the fictitious o o o o o o o sWOM sender] to last for a long time If you read this, tick off o o o o o o o strongly disagree I feel strongly linked to o o o o o o o [name of the fictitious sWOM sender]. The relationship with [name o o o o o o o of the fictitious sWOM sender] is important to me.

What is your opinion on [name of the fictitious sWOM sender] with respect to movies? [name of the fictitious sWOM sender] is …

-3 -2 -1 0 +1 +2 +3 Dishonest o o o o o o o Honest Unreliable o o o o o o o Reliable Untrustworthy o o o o o o o Trustworthy Not an expert o o o o o o o Expert Inexperienced o o o o o o o Experienced Unknowledgeable o o o o o o o Knowledgeable Unqualified o o o o o o o Qualified Unskilled o o o o o o o Skilled

Appendix To what extent do you agree with the following statement?

Strongly Disagree Somewhat Neither Somewhat Agree Strongly disagree disagree disagree, agree agree nor agree If I were to choose a movie, I would consider o o o o o o o Dirty.

To what extent do you agree with the following statements? A

-3 -2 -1 0 1 2 3 I am likely to say I am likely to say positive negative things about o o o o o o o things about Dirty to Dirty to other people. other people.

I am not likely to I am likely to recommend Dirty to a o o o o o o o recommend Dirty to a friend or colleague. friend or colleague.

Appendix

I am likely to discourage I am likely to encourage friends and relatives to o o o o o o o friends and relatives to go see Dirty. go see Dirty.

312

Appendix 3 – Alternative analyses

3A. Linear regression with actual personalization (Chapter 3)

Study 1 Perceived Relevance Brand Attitude Click intention B Sig. B Sig. B Sig. Constant 2.217 <.001 2.532 <.001 .729 <.001 Actual personalization (1: personalized) -.127 .578 -.064 .653 -.133 .345 Attitude toward Facebook .030 .643 .135 .041 Actual personalization * attitude toward Facebook -.229 .069 -.199 .148 Perceived Relevance .334 <.001 .436 <.001 Gender .123 .618 .251 .138 -.011 .941 R² .004 .296 .393 Study 2 Perceived Relevance Brand Attitude Click intention B Sig. B Sig. B Sig. Constant 2.053 <.001 3.105 <.001 .917 <.001 Actual personalization (1: personalized) -.680 .002 -.327 .046 -.083 .546 Attitude toward Facebook .019 .778 .106 .165 Actual personalization * attitude toward Facebook .111 .395 .099 .511 Perceived Relevance .259 .001 .414 <.001 Gender .043 .872 -.149 .488 -.145 .494 R² .066 .181 .304

Appendix

313

314 3B. Linear regression with actual personalization (Chapter 4 – Study 1) Appendix

Perceived Relevance Perceived Intrusiveness Brand Attitude Click intention

Constant 1.526 <.001 2.851 <.001 2.609 <.001 1.568 .002 Actual personalization (gender) -.162 .349 .490 .003 -.068 .531 -.290 .167 Actual personalization (age) -.042 .807 .306 .091 .012 .913 -.131 .556 Actual personalization (interests) -.118 .483 .363 .034 .023 .835 -.284 .189 Perceived relevance .277 <.001 .565 <.001 Perceived intrusiveness -.183 .003 -.249 .006 Gender .011 .940 -.149 .248 -.047 .584 -.099 .517 Age -.158 .018 -.030 .673 .040 .295 .005 .943 R² .041 .056 .253 .262

3C. Linear regression with actual personalization (Chapter 4 – Study 2)

Perceived Perceived Brand Attitude Click intention WOM intention Relevance Intrusiveness Constant 3.599 <.001 3.653 <.001 4.469 <.001 1.757 .013 3.602 <.001 Actual personalization (gender + age) -.171 .549 -.054 .873 -.032 .882 .028 .900 .083 .701 Actual personalization (gender + age + interests) .344 .263 -.041 .911 -.604 .011 .053 .820 .085 .712 Actual personalization (gender + age + interests + life- .353 .288 -.392 .275 -.087 .697 .124 .615 .377 .101 event) Actual personalization (gender + age + interests + life- .426 .171 -.139 .696 .053 .801 .540 .047 .031 .903 event + friend referrals) Perceived relevance .352 <.001 .588 <.001 .234 <.001 Perceived intrusiveness -332 <.001 -.225 <.001 -.093 .061 Gender -.027 .893 -.119 .579 -.086 .517 -.053 .745 -.002 .987 Age -.025 .277 .039 .117 .006 .742 -.018 .359 -.010 .572 R² .027 .018 .451 .460 .157

3D. Linear regression with actual message valence, tone of voice and service type (Chapter 6)

Att service provider PI WOM Constant 3.837 <.001 3.613 <.001 3.706 <.001 Actual valence (1: positive) 1.807 <.001 1.442 <.001 .880 <.001 Actual tone of voice (1: emotional) .119 .286 .212 .106 -.110 .294 Actual service type (1:bar) .479 <.001 .790 <.001 .464 <.001 Valence x tone of voice -.657 .003 -.442 .093 -.487 .020 Valence x service type 1.037 <.001 1.101 <.001 .832 <.001 Tone of voice x service type -.056 .802 .335 .202 -.273 .193 Valence x tone of voice x service type .374 .401 -.069 .896 .381 .363 R² .448 .320 .232

Appendix

315

Nederlandse Samenvatting

S Nederlandse samenvatting

Over de hele wereld hebben mensen sociale media omarmd als een belangrijk communicatiekanaal: in Appendix april 2019 waren er wereldwijd 3.48 miljard sociale media gebruikers (Kemp, 2019). Sociale media, en

sociale netwerksites is het bijzonder, zijn belangrijk geworden voor zowel hun persoonlijke als hun sociale leven. Sociale media worden vooral gebruikt omwille van amusement, om tijd te doden en sociale activiteiten (Ku et al., 2013). Ook adverteerders hebben hun weg gevonden naar deze platformen en proberen in contact te treden met hun (potentiële) klanten via merkcommunicatie. Zo rapporteerde

Facebook een reclame-inkomst van meer dan 50 miljoen dollar in 2018 (Facebook Inc., 2019c). De voornaamste doelstelling van dit proefschrift is om de reacties van consumenten op merkcommunicatie op sociale netwerksites te bestuderen (SNSs). We hanteren hierbij twee invalshoeken: gepersonaliseerde reclame en mond-tot-mondreclame.

Door het gebruik van sociale netwerksites geven gebruikers een grote hoeveelheid persoonlijke informatie vrij, zoals hun locatie, hun interesses, hun demografische gegevens, etc., waarmee een digitale vingerafdruk gecreëerd wordt. Al die informatie kan gebruikt worden door adverteerders om hun reclameboodschappen aan te passen aan specifieke targets (Kelly et al., 2010; Sundar & Marathe,

2010). Facebook Inc. (2019a), bijvoorbeeld, biedt de mogelijkheid om specifieke targets te specifiëren op basis van locatie, demografische gegevens, interesses, connecties en gedrag. Gepersonaliseerde reclame wordt doorgaans als effectiever beschouwd dan niet-gepersonaliseerde reclame wat betreft visuele aandacht (Bang & Wojdynski, 2016; Pfiffelmann et al., 2019), attitudes en de gedragsintenties van consumenten (Li & Liu, 2017).

Daarnaast bevat merkcommunicatie op sociale netwerksites ook mond-tot-mond reclame (sWOM), een tweewegs-communicatiemiddel. Een belangrijk deel van het moderne leven is het online delen van ervaringen met producten en services (Kim, Jang, et al., 2015). Sociale netwerksites stellen mensen in staat om informatie te verspreiden naar een groot aantal mensen en instellingen door interpersoonlijke online interacties (boyd & Ellison, 2007; Hennig-Thurau et al., 2004). Door hun populariteit bezoeken consumenten vaak sociale netwerksites om informatie op te zoeken over bedrijven of reviews te lezen over producten (Shaw, 2018). Onderzoek naar sWOM toont aan dat deze informatie een belangrijke rol kan spelen in het vormen van attitudes en gedragingen. (e.g., Karakaya & Barners, 2010; Rui et al.,

2013; Schivinski & Dabrowski, 2015; Wang et al., 2012).

318 Nederlandse samenvatting

De eerste doelstelling van het proefschrift was om de relatieve belangrijkheid van verschillende personalisatie-elementen te bepalen bij het uitlokken van de perceptie van personalisatie. Voorgaand onderzoek heeft aangetoond dat werkelijke en gepercipieerde personalisatie niet automatisch gelijk zijn (Kramer et al., 2007; Li, 2016). Het onderzoek in hoofdstuk 2 toont aan dat alle elementen in het onderzoek (interesse, locatie, leeftijd, levensgebeurtenissen, geslacht en verwijzingen door vrienden) de perceptie van personalisatie kunnen uitlokken. Interesses blijken het meest belangrijk. Deze resultaten blijven stabiel overheen productkenmerken en demografische kenmerken van de respondent. Voorgaand onderzoek rapporteerde inderdaad dat het gebruik van interesse leidt tot een succesvolle manipulatie van personalisatie (Kalyanaraman & Sundar, 2006; Li, 2016).

De tweede doelstelling was om de positieve en negatieve effecten van gepersonaliseerde reclame op de attitude en gedragsintenties van consumenten te bestuderen. Daarmee speelt dit proefschrift in op de kenniskloof die er bestaat omtrent de verwerkingsmechanismen van gepersonaliseerde reclame op sociale netwerksites. In Hoofdstuk 3 en 4 werd een sterk positief indirect op merkattitude, klikintentie en positieve mond-tot-mond intentie via persoonlijke relevantie gevonden.

Hiermee bevestigt dit proefschrift voorgaand onderzoek (e.g., Aguirre et al., 2015; Bleier & Eisenbeiss,

2015; Kim & Huh, 2017). Hoofdstuk 4 voegt gepercipieerde opdringerigheid en gepercipieerd entertainment toe, naast persoonlijke relevantie. In tegenstelling tot onze verwachtingen werd een positief, indirect effect gevonden van gepercipieerde personalisatie op merkattitude, klikintentie en positieve mond-tot-mond intentie via gepercipieerde opdringerigheid. Gebaseerd op van Doorn and Hoekstra (2013) en White et al. (2008), verwachtten we dat het gebruik van persoonlijke informatie het gevoel van opdringerigheid zou verhogen. Onze bevindingen wijzen op het tegenovergestelde. Dat kan te wijten zijn aan het feit dat de informatie in de advertentie waardevoller is waardoor de reclameboodschap minder opdringerig is dan de niet-gepersonaliseerde reclameboodschap (e.g., Ketelaar et al., 2018; Kim & Han, 2014). In lijn met voorgaand onderzoek vonden we een positief indirect effect via gepercipieerde entertainment. Kim and Han (2014) stellen dat een gepersonaliseerde S boodschap de behoeften van de consumenten beter weerspiegelt en dat verhoogt, op zijn beurt, de perceptie van entertainment. Bovendien worden deze advertenties minder vaak vermeden (Edwards et al., 2013).

319 Nederlandse samenvatting

Hoofdstuk 4 voegt ook nog twee tweede-orde mediatoren toe aan het verwerkingsmodel: self-brand Appendix connection en reactance tegenover de advertentie. Concreet heeft gepercipieerde personalisatie een

positief effect op self-brand connection via persoonlijke relevantie en entertainment, maar niet via opdringerigheid. Dit wordt verklaard via het self-referencing proces: zowel via persoonlijke relevantie als via entertainment kunnen de psychologische behoeften van de consumenten vervuld worden. Self- brand connection leidt, op zijn beurt, tot een positievere merkattitude, klikintentie en positieve mond- tot-mondreclame intentie. Uit onze resultaten blijkt ook dat gepercipieerde personalisatie reactance tegenover de advertentie kan verlagen via gepercipieerde entertainment, relevantie en opdringerigheid.

Dit ligt in lijn met de bevindingen van Edwards et al. (2013): gepercipieerd entertainment kan psychologische reactance tegenover een gepersonaliseerde reclameboodschap doen dalen door de meerwaarde die de boodschap biedt aan consumenten. In tegenstelling tot onze verwachtingen heeft reactance tegenover de advertentie geen significant effect op consumentenreacties. Dit kan verklaard worden door de bevindingen van Marotta, Zhang, and Acquisti (2015): ondanks dat consumenten het niet leuk vinden om online getraceerd te worden, appreciëren ze vaak toch de meerwaarde van online gepersonaliseerde reclame. Bijgevolg zijn ze minder geneigd zich te verzetten tegen de reclameboodschap. Hoofdstuk 5 focust op het relatieve belang van positieve (voordelen) en negatieve

(kosten) effecten van gepercipieerde personalisatie. Onze resultaten geven aan dat het (verwachte) positieve effect van persoonlijke relevantie minder sterk doorweegt dan gepercipieerde creepiness. Zelfs wanneer er slechts 1 personalisatie-element (de verwijzingen door vrienden) wordt gebruikt, kunnen consumenten de reclameboodschap als eng bestempelen en in dat geval heeft persoonlijke relevantie geen significant effect.

De derde doelstelling van dit proefschrift was om randvoorwaarden van de effecten van gepersonaliseerde reclame op de attitude en de gedragsintenties van consumenten, alsook de geassocieerde verwerkingsmechanismen, te bestuderen. Hoofdstukken 3 en 4 onderzoeken de invloed van de attitude tegenover de sociale netwerksite. De resultaten van hoofdstuk 3 tonen aan dat voor producten waarbij de consumenten hoog betrokken zijn, de attitude tegenover de sociale netwerksites de relatie tussen gepercipieerde personalisatie modereert en merkattitude of klikintentie niet. Voor een product waarbij consumenten weinig betrokken zijn, modereert de attitude het effect tussen gepercipieerde personalisatie en klikintentie (maar niet merkattitude). Het positieve effect van

320 Nederlandse samenvatting gepercipieerde personalisatie was enkel significant voor consumenten met een matige tot positieve attitude tegenover de sociale netwerksite. In hoofdstuk 4 wordt onderzocht of attitude tegenover de sociale netwerksite de indirecte effecten via persoonlijke relevantie en gepercipieerde opdringerigheid op consumentenreacties beïnvloeden. De resultaten tonen aan dat dit niet het geval was. Echter, het directe effect van percipieerde personalisatie op positieve mond-tot-mondreclame en klikintentie werd, net zoals in hoofdstuk 3, gemodereerd door de attitude tegenover de sociale netwerksite. Meer bepaald stijgt het positieve effect van gepercipieerde personalisatie naargelang de attitude tegenover de sociale netwerksite positiever wordt.

Hoofdstuk 4 onderzoekt ook de invloed van de modererende rol van gepercipieerde privacybescherming door de sociale netwerksite. Onze resultaten tonen aan dat het indirecte effect via persoonlijke relevantie op klikintentie (maar niet merkattitude of positieve mond-tot-mondreclame) inderdaad gemodereerd wordt door gepercipieerde privacybescherming door de sociale netwerksite. Meer bepaald, het positieve indirecte effect via persoonlijke relevantie op klikintentie wordt sterker naarmate de perceptie van privacybescherming door de sociale netwerksite hoger positiever wordt. Dit kan te wijten zijn aan het feit dat gepercipieerde privacycontrole een negatief effect heeft op privacybezorgdheid en, indirect, op advertentievermijding. Als consumenten overtuigd zijn dat hun privacy wordt beschermd, zou het kunnen dat ze de gepersonaliseerde advertentie meer appreciëren, waardoor ze de persoonlijke relevantie ervan meer waarderen.

Hoofdstuk 5 onderzoekt de modererende rol van het type van bron (i.e., de adverteerder). De negatieve impact van personalisatie via gepercipieerde creepiness werd enkel gevonden voor een gezondheidswebsite, een overheidswebsite en een online krant, maar niet voor commerciële websites.

Een vierde, en laatste, doelstelling van dit proefschrift was om randvoorwaarden te onderzoeken van het valentie-effect van sWOM boodschappen op de attitudes en gedragsintenties van consumenten. Hoofdstuk 6 onderzoekt het modererende effect van tone of voice (feitelijke of emotioneel) en type van service (utilitair of hedonistisch). Onze resultaten geven aan S dat het effect van valentie sterker is voor feitelijke boodschappen, vergeleken met emotionele.

Bovendien is het effect van valentie sterker voor hedonistisch services vergeleken met utilitaire services.

Tot slot verwachtten we dat het laten overeenstemmen van de tone of voice met het service type zou

321 Nederlandse samenvatting

leidden tot een grotere impact op consumentenreacties. Onze resultaten tonen echter aan dat zowel Appendix voor utilitaire als voor hedonistische services, het informatieve effect van een feitelijke boodschap

sterker doorweegt dan het potentiële matching-effect .

Hoofdstuk 7, tot slot, onderzoekt het modererende effect van interpersoonlijke en persoon-tot-site relationele kenmerken op het effect van valentie op de gedragsintentie en positieve mond-tot- mondreclame intentie van consumenten. Uit het onderzoek blijkt dat interpersoonlijke homophily en brongeloofwaardigheid het effect van valentie op gedragsintentie en positieve mond-tot-mondreclame intentie versterken. Als enkel de persoon-tot-site relationele kenmerken worden beschouwd, dan versterken ook de persoon-tot-site homophily en brongeloofwaardigheid het effect van valentie op gedragsintentie en positieve mond-tot-mondreclame intentie. Echter, wanneer zowel de interpersoonlijke als de persoon-tot-site variabelen worden opgenomen, worden de effecten van de persoon-tot-site relationele kenmerken niet-significant.

Het concluderende hoofdstuk bespreekt naast de theoretische en praktische implicaties een aantal beperkingen van de uitgevoerde studies, en reikt ideeën aan voor verder onderzoek.

322