Master Thesis MSc Business Administration-‐ Marketing track
Final draft
Personalized advertising: Perceived privacy concerns on different social
media channels
Supervisor: Jonne Guyt
By: Fayrouz Salem-‐ 11143606
24
thof June 2016
ABSTRACT
The access and integration of consumers’ personal information has become imperative for today’s marketing practices. One of the new realities in advertising is that consumers’ personal information, that is available on social media networks, can be used for targeting purposes. By doing so, consumers are provided with advertisements that fit their preferences and characteristics. Nevertheless, the use of personal information sets up a trade-‐off between the relevance of advertising and privacy concerns. This study attempts to examine the perceived privacy concerns of personalized advertising on social media platforms by conducting an experiment. The findings reveal that the individuals who were exposed to a personalized advertisement had higher perceived privacy concerns than those exposed to a generic advertisement. However, the results also showed that this effect was not moderated by different social media platforms.
Keywords: personalized advertising, privacy, social media platforms
STATEMENT OF ORIGINALITY
This document is written by student Fayrouz Salem who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than the sources mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business Economics is responsible solely for the supervision of completion of the work, not for contents.
TABLE OF CONTENTS
1.
INTRODUCTION……….5
2.
LITERATURE REVIEW……….10
2.1 Personalized advertising
………10
2.2 Personalized advertising on social media
………..12
2.3 Privacy concerns
………...14
2.4 Differences in platform
………15
2.5 Conceptual framework
………20
3.
RESEARCH DESIGN AND METHODOLOGY……….21
3.1 The sample
………..21
3.2 Research design
……….21
3.3 The procedure
………22
3.4 Variables & measurements
……….23
4.
RESULTS...24
4.1 Preliminary analysis
……….24
4.2 Manipulation check
………..24
4.3 Formal test of the model
………..27
4.4 Likeability & click rate………....
29
5.
DISCUSSION & CONCLUSION……….32
6.
LIMITATIONS & FUTURE RESEARCH……….38
7.
REFERENCES……….39
1. INTRODUCTION
Over the past years, social media has become an increasingly popular phenomenon attracting millions of users, whom have made the use of such networks a part of their daily practice (Ellison, 2007). Social networking sites (SNS) can be defined as “web-‐ based services that allow individuals to (1) construct a public or semi-‐public profile within a bounded system, (2) articulate a list of other users with whom they share a connection, and (3) view and traverse their list of connections and those made by others within the system” (Ellison, 2007, p. 211).
Social media platforms are nowadays not only serving the purpose of connecting with friends and family, but have also become an integral part of the marketing mix (Barbu, 2014). Social media as a marketing tool encompasses “a wide range of online, word-‐of-‐mouth forums including blogs, company sponsored discussion boards and chat rooms, consumer-‐to-‐consumer e-‐mail, consumer product or service ratings websites and forums, Internet discussion boards and forums, moblogs (sites containing digital audio, images, movies, or photographs), and social networking websites, to name a few” (Mangold & Faulds, 2009, p. 358). The presence of social media has given companies the possibility to use new tools and strategies to interact and communicate with consumers (Mangold & Faulds, 2009). Firms exhaust social media platforms to communicate information about their products and services by placing advertisements. With the increasing popularity of social media, global social network ad spending is also accelerating and was predicted to reach $25.14 billion in 2015 (“Social Network Ad Revenues”, 2015). Online advertising nowadays constitutes a large proportion of the advertising market with big players such as Google and Facebook relying primarily on Internet advertising to generate revenue (Goldfarb, 2014). In fact, Facebook was expected to capture $16.29 billion in ad revenues worldwide in 2015, which accounts
for 64,8% of the total social media ad spending (“Social Network Ad Revenues”, 2015). Advertising through social networking sites has not only changed the advertising domain by cutting down costs, but it also revolutionized the way advertisers approach consumers (Gangadharbatla, 2008). The main benefit of advertising through social media is that companies have access to the personal information provided by users, which can be used for targeting purposes (Luna-‐Nevarez & Torres, 2015). The development of social media and access to a large database of consumer personal data, create new research challenges in understanding how consumers respond to social media advertising.
When users create an account on Facebook they are required to fill in demographic information such as their first and last name, their email address, their gender and date of birth (Barbu, 2014). The database of users’ personal information provided on social media platforms facilitates the personalization of ads which is defined by Montgomery & Smith (2009, p. 130) as “the adaptation of products and services by the producer for the consumer using information that has been inferred from the consumer's behavior or transactions”. Even though much of the existing research elaborates on the effects of personalization, little is known about how consumers perceive personalization on different social media platforms. Existing research suggests that personalization on social media can be effective as consumers are shown ads that match their interest (Lambrecht & Tucker, 2013). On the other hand the collection and usage of consumers’ data have aroused concerns about consumer privacy (Goldfarb, 2014). For those who are concerned with their privacy, personalized advertising may even trigger negative responses such as ad avoidance. In fact, 86% of young adults do not prefer personalized ads if it means that their browsing behavior on websites is being tracked (Turow et al., 2009). However, these findings do not account
for personalized advertising on social media platforms therefore leaving a literature gap in marketing. Furthermore with more social media platforms introducing the concept of personalized advertising, it becomes interesting to investigate whether consumers perceive any differences regarding privacy concerns on different social media platforms. It is important for marketers to understand which kind of information consumers can tolerate and therefore what the most effective way is to approach consumers on social media platforms without intruding their privacy.
This study aims to bridge the existing literature gap by conducting an experiment that investigates how consumers respond to personalized ads on different social media platforms. The scope of the current study focuses on Facebook and Instagram, which are two of today’s biggest social media networks therefore the main research question is: How do different social media platforms (i.e. IG/FB) moderate the effect of personalized ads on perceived privacy concerns?
Note: Throughout the study the terms social media platforms and social networking sites will be used indifferently.
Practical Contribution
With social media networks accounting for one-‐third of all online display advertising, it is important for marketers to understand how it can be used in the most effective manner to target consumers (Tucker, 2014). Personalization has numerous benefits for both marketers and online users, however these benefits do not always materialize (Stockman, 2010). From a practical point of view it is imperative to further understand which platforms are more effective for personalized advertising without evoking the feeling that consumers’ privacy is being intruded. If advertising on social media
platforms is being avoided due to perceived privacy concerns then the antecedents of these privacy concerns need to be understood. It is also important to whether avoidance is due to the nature of the advertisement (generic/personalized) or whether it is due to the platform. For instance are privacy concerns higher on Facebook than on Instagram? This is relevant for marketers who wish to advertise effectively on social media platforms in order to persuade consumers to buy their products and services.
Theoretical Contribution
Despite the growing importance of personalized advertising on social media within the realm of marketing, to date there is no empirical study that investigates the effectiveness of personalized advertising on different social media channels. There is also no academic work on how these different social media networks moderate the effect of personalized ads on perceived privacy concerns. Personalized advertising on social media is a fairly new area in marketing with many literature gaps that can be partially bridged with this research (Hadija et al., 2012). This research is of theoretical relevance because it contributes to the online advertising literature that examines personalization on social media platforms. Furthermore it contributes to the extant literature about privacy concerns regarding the disclosure of consumer personal information, privacy attitudes and privacy behaviour. Therefore, academic relevance of this study is twofold: first it will unravel consumers’ perceived privacy concerns regarding personalized advertising on different social media networks. Second it will further illuminate consumers’ online privacy behavior.
Thesis overview
The remainder of the study is structured in the following chapters. First, a review of the extant literature about personalized ads, perceived privacy concerns and social media will be provided in chapter two. Within the literature review the hypotheses will be presented followed by the conceptual framework of the study. In chapter three the research design and methodology will be explained. Next, the results of the study will be presented and thoroughly analyzed in chapter four. Finally, in chapter five the empirical findings will be discussed and conclusions will be drawn. Hereafter, the limitations of the study will be highlighted and recommendation for future research will be given in chapter six.
2. Literature review
This section provides a comprehensive review of the existing literature that covers the key concepts relevant to this study. The purpose of this review is to analyze what has already been studied with regards to personalized ads on social media platforms and privacy concerns. First, literature on personalized advertising will be reviewed. Next, previous research about the effectiveness of social media advertising will be delved into. Thereafter, current online privacy concerns will be examined followed by a section that highlights the differences between Facebook and Instagram.
2.1 Personalized Advertising
Over the past few years the effectiveness of traditional advertising, which consists of identical messages and targets a mass audience, has been questioned (Yu et al., 2009). In contrast, as new technologies have developed, companies have shifted the focus of their effort from generic advertising to online personalized advertising which uses individual’s personal information to deliver a customized message at the right time to the right person (Yu et al., 2009). Individual’s personal information may include one’s personal email address, name, residence or even personal information such as where they have shopped, browsed websites and preference for a specific product or even one’s hobby (Yu et al., 2009). According to Dijkstra (2008) persuasive information can be tailored to individual characteristics through the use of advanced computer technology. From a psychological perspective, Dijkstra (2008) claims that personalization can be an effective method of persuasion because by mentioning for instance one’s name, which increases involvement by making the content information personally relevant, it can increase the central processing. The benefit of personalized advertising is that it enables a quick focus on customers’ desires therefore by
communicating only relevant information; their search efforts are minimized (Srinivasan et al., 2002). Accordingly, Pavlou and Stewart (2000) suggest that when consumers receive messages that are relevant to them, this can result in more purchase intentions and other responses that are deemed desirable to the company. Consequently, many Internet firms are nowadays collecting personal data from their users and are using this data to target consumers on an individual level (Tucker, 2014).
Furthermore, advanced technology enables firms to process historic browsing data on an individual level which allows them to offer personalized recommendations to consumers who return to the website (Lambrecht & Tucker, 2013). In addition, browsing history data is also used to advertise content on external websites meaning that: consumers are shown ads of products that they have previously viewed on a firm’s website when they are browsing websites not related to the products they viewed (Lambrecht & Tucker, 2013). Before the study by Lambrecht & Tucker, (2013) there was little empirical evidence about whether personalized product recommendations are effective when they are displayed on external websites in comparison to ads displayed internally on a firm’s website. Their findings suggest that dynamic retargeting, which is when consumers are shown ads specific to the product they were previously viewing but did not purchase, is effective in terms of persuading the consumer to purchase the product only when the consumers are actively browsing product information such as review sites. In contrast, when consumers are not searching for product specific information and their preference level is construed at a high level, meaning they only have a broad idea of what they prefer, then generic ads are more effective (Lambrecht & Tucker, 2013). Other studies have also investigated the effectiveness of personalized advertising. Howard and Kerin (2004) found that in general the response rate to advertisements that included personal information was higher than the response rate to
advertisements that did not include such information. Extending on the effectiveness of personalized advertising, a study by Urban et al (2013) demonstrated that the implementation of morphing online banners was effective because it “enables a website to learn, automatically and near optimally, which banner advertisements to serve to consumers to maximize click-‐through rates, brand consideration, and purchase likelihood” (Urban et al, 2013, p.2). In fact, the online banners that were personalized, in the sense that they matched consumer’s cognitive styles and buying process, almost doubled click-‐through rates relative to the control banners (Urban et al, 2013).
2.2 Personalized advertising on social media
Personalized online advertising has gained momentum and even reached social networking sites which are platforms used for information sharing, video sharing, photo sharing and blogging (Gangadharbatla, 2008). The core benefit of using social media platforms for advertising purposes is that firms can utilize the personal information provided by users such demographics, to create personalized advertisements that target individual consumers (Luna-‐Nevarez & Torres, 2015). Targeting consumers through social media platforms may increase the chance that consumers will receive relevant advertising messages therefore reducing the chance of annoyance and frustration associated with advertisements (Stockman, 2010). Advertising has become the primary source of revenue for most social networking sites and they are generally designed to target the individual user (Gangadharbatla, 2008). Facebook, one of the largest social media platforms, has incorporated the use of personalized advertisements as it main advertising strategy by including retargeted ads on its users’ news feed (Rusli, 2013). By using Facebook as a medium, advertisers can collect demographic information about the users and automatically match ads to a specific audience (Lambrecht & Tucker, 2013).
Recently, Instagram, which is one of the largest social media platforms for sharing photos, introduced the possibility for companies to place sponsored ads (“Instagram Advertising”, 2016). Instagram has a community of more than 400 million users therefore making it one of the world’s largest mobile advertising platforms (“Instagram Advertising”, 2016). The advertisements that appear on one’s personal Instagram account are targeted ads that use data from its parent company Facebook (Griffith, 2015). According to a study conducted by Instagram, advertisers on the platform are seeing positive results such as an increase in mass awareness, an increase in website sales and mobile app downloads (“Instagram Advertising”, 2016).
Social networking users are nowadays frequently exposed to advertisements when using social media platforms therefore given this, it becomes relevant to further understand consumer attitudes towards this type of advertsing (Luna-‐Nevarez & Torres, 2015). Despite the growing importance of social media advertising, there is little empirical evidence that explains the factors that drive consumer’s attitudes towards these ads. There is a clear literature gap regarding how consumers respond to personalized ads on social media, which needs to be filled by empirical research.
On the one hand it may be argued that personalized ads are more appealing to consumers because they are more in line with their interests (Tucker, 2014) but on the other hand this type of advertising may also be perceived as an intrusion on privacy (Stone, 2010). In fact, personalization of ads can lead to reactance, which is a motivational state in which consumers behave in the opposite way to that intended (White et al. 2008). As argued by White et al. (2008, p. 40) “reactance occurs when highly personalized messages lead consumers to feel constrained by the sense of being too identifiable or observable by the firm”. Given this, the privacy concerns related to personalized advertising should be taken into account.
2.3 Privacy concerns
While personalization generates several benefits for both consumers and marketers, it is not without drawbacks. The information-‐rich environment facilitated by the Internet has enabled firms to easily collect and store data about consumers however, the collection of this data, has also resulted in the rise of concerns regarding consumer privacy (Goldfarb, 2014). Consumers main concerns with regards to the collection of personal data is how much data is collected, what type of information is collected, how it is used without their knowledge and consent and whether this is freely shared with other marketers (Peltier, 2009). According to Goldfarb and Tucker (2011) consumers who are concerned about their privacy react negatively to personalized ads. A dilemma exists in which on the one hand personalized ads might be more appealing to consumers because it is aligned with their preferences but on the other hand consumers may be concerned with the use of their personal information. Turow et al. (2009) find that when personalized ads are a result of following consumers’ behavior on websites other than the ones they have visited, 86% of the young adults would rather not have personalized ads. However, according to Culnan & Armstrong (1999, p, 106) “individuals are less likely to perceive information collection procedures as privacy-‐invasive when (a) information is collected in the context of an existing relationship, (b) they perceive that they have the ability to control future use of the information, (c) the information collected or used is relevant to the transaction, and (d) they believe the information will be used to draw reliable and valid inferences about them”. In a more recent study, Tucker (2014) suggests that when advertising is perceived by consumers as intrusive, social networking websites can resolve this by giving users control over how their information is used. Social media platforms such as Facebook have already considered this solution by introducing policy changes that allow users to easily control the information they want to automatically be
displayed and also gave users control over the tracking of their personal data by third parties (Tucker, 2014). A surprising finding in the study conducted by Tucker (2014) is that personalized advertising became more effective when consumers were enabled to control their privacy on Facebook. This finding suggests that click-‐through rates will increase when consumers are reassured control of their privacy (Tucker, 2014).
Despite giving users control of their privacy, in general personalization has generated criticism as consumer information is gathered therefore constituting an invasion on consumer’s privacy (Montgomery & Smith, 2009). As a consequence, highly personalized messages can result in consumers feeling constrained by the idea of being too identifiable or constantly being observed by firms (White et. al 2008). Prior to the introduction of personalized ads, generic ads were the norm, which are targeted at a mass audience. Additionally, due to the lack of literature a preliminary study (Appendix I) was conducted among a group of 7 individuals to gain introductory insights. Participants were asked to share their opinion regarding personalized advertisements on social media networks. The study revealed that overall personalized ads are perceived as more privacy intruding than generic ads. Given the use of personal information in personalized ads, the following hypothesis is proposed:
H1: Personalized ads generate higher perceived privacy concerns than generic ads
2.4 Differences between platforms
While there are ample types of social media platforms with different functional properties, the study will focus on Facebook and Instagram, being two of the most widely used social networks. “Facebook is a social media platform that allows users to create profiles and become “friends” with other users. Friends are able to communicate and share videos, pictures, and links with each other through “status updates” and private
messages. Users can also create groups, “like pages,” and event pages” (Al-‐Bahrani & Patel, 2015, p.61). In contrast, Instagram is a photo sharing community that “allows users to share pictures or videos that are 15 seconds or less with their networks” (Al-‐Bahrani & Patel, 2015). While both platforms have similar social aspects of engagements: liking and commenting on a post, the two platforms differ in many other aspects, which will be highlighted in this section.
The biggest difference between Facebook and Instagram is that from the offset, Instagram was introduced as a mobile application while Facebook was initially a desktop social networking website (Goggin, 2014). However, with today’s mobile revolution, Facebook is tapping into this trend: “as a mobile-‐first, mobile-‐best platform with 21m mobile visitors every day, Facebook is perfectly positioned at the centre of this seismic shift,” said Mendelsohn. “With one in five mobile minutes spent with Facebook and Instagram, and people checking their News Feed up to 14 times a day, Facebook is part of the connective tissue of the mobile web.” (“The Mobile Revolution”, 2014). In fact, eMarketer reported that mobile Internet usage is close to surpassing traditional desktop Internet usage (“Digital Set to Surpass TV”, 2014). With both Facebook and Instagram tapping into the smartphone trend, the difference between the two platforms is minimized. Despite this, there are still numerous differences to consider.
Firstly, registration on Facebook and Instagram differ in terms of the degree of personal information that is disclosed. Facebook is an online community, which only permits registration using authentic identities and does not allow individuals to maintain more than one personal account (www.facebook.com). When signing up, individuals are required to provide their first and second name, telephone number, date of birth and gender (Barbu, 2014). Users must also agree to the terms of service, in which it is stated that Facebook has the right to collect users’ demographic information (Wilson et al.,
2012). In contrast, Instagram does not require one to provide their gender, age or telephone number. Additionally, individuals are permitted to have multiple Instagram accounts with usernames rather than official names (www.Instagram.com).
Furthermore, while Instagram is mainly a photo-‐sharing platform, Facebook’s features are not limited to posting photos. Facebook for instance allows users to “check in” at places and to tag specific locations on individual status updates allowing other users to know where they are localized at a specific time (Chang et al., 2014). Facebook also gives users the opportunity to share personal information such as current employment, date of birth, places one has lived in, relationship status and so forth (Gangadharbatla, 2008). It may be argued that Facebook is a platform that provides individuals with the opportunity to present themselves authentically and in a positive light to other users in their Facebook friends list (Wilson et al., 2012). However, by presenting such information on Facebook, individuals may be prone to face privacy harms such as identify theft or less drastically: have strangers know about their life history (Acquisti et al., 2015).
As for the nature of friends, to connect and fully view other users’ Facebook accounts, a friend request has to be sent which users can either accept or decline Gangadharbatla, 2008. Once a friend request has been accepted, users become “friends” and can view each other’s posted content. In fact, Ellison et al. (2007) argue that the most common internal motivation for using Facebook’s is to maintain interpersonal relationships and friendships irrespective of time and physical space. According to Wilson et al., (2012) the majority of the Facebook relationships are also offline relationships. In contrary, Instagram users can connect with other users by following their account; in fact, Instagram accounts are public by default (Bakhsi et al., 2014) which allows any user on the platform to freely view content and to follow an account of interest without the acceptance of a friend request. It may therefore be argued that the
relationship between Facebook friends is more personal than users following each other on Instagram.
Table 1. Overview of characteristics that make the platform more personal Instagram < Facebook
Registration + ++
Nature of friends + ++
Account privacy settings + ++
The characteristics of Facebook have shown that it is a platform that encourages the sharing of personal information “But the sharing of content and personal information on Facebook comes with certain potential privacy risks, including unintentional disclosure of personal information, damaged reputation due to rumors and gossip, unwanted contact and harassment, vulnerability to stalkers or pedophiles, use of private data by a third party, hacking, and identity theft” (Wilson et al. 2012 p, 212). Moreover, a study by Christofides et al., (2009) confirmed that overall, participants were concerned with their personal privacy on Facebook. In fact, Acquisti et al., (2004) argues that in general, individuals are concerned about the privacy and security of their personal information that they share online because the security of their personal information is not guaranteed.
However, a discrepancy exists between the disclosure of personal information and privacy concerns as was revealed in a study by Acquisti et al., (2006) in which participants who reported to be “very worried” about strangers finding out where they lived but still revealed this information on their profile. This discrepancy can be explained by what is known today as the privacy paradox: “privacy concerned individuals are willing to trade-‐off privacy for convenience or to bargain the release of very personal
result suggests that even when people are concerned with their privacy on Facebook they might still be willing to share their personal information.
Additionally, in the light of the absence of academic work regarding personalized advertising on different social media platforms, the preliminary study (Appendix I) helped gain some insights into whether users perceive any differences in terms of privacy concerns. Several differences were revealed: overall personalized ads on Instagram were better perceived than on Facebook and were regarded as more relevant. According to Luna-‐Nevarez & Torres (2015) advertisements on social media are more likely to be avoided if they are regarded as irrelevant. This finding in conjunction with the result of the preliminary study suggests that Instagram advertisement may be perceived as more relevant than advertisements on Facebook. Furthermore, the participants were asked about their privacy concerns and surprisingly, personalized ads on Facebook intruded privacy more than the personalized advertisements on Instagram. One explanation given by a participant was the following: “I feel that Facebook intrudes my privacy more as it is a 'closed medium' or my Facebook is at least. Instagram feels more open as it is a medium where u would like to be seen beyond your network due to the hashtags etc-‐. You are more aware of the exposure”. One of the differences between Facebook and Instagram is that Instagram accounts are public by default whereas Facebook accounts are private. Additionally, more personal information is provided when signing up for a Facebook accout compared to an Instagram account. Furthermore, friends on Facebook are usually individuals in one’s offline network whereas followers on Instagram may not be. All in all, Facebook is a platform where more personal details are shared than on Instagram which can be used for advertising purposes therefore the following hypothesis is suggested: H2: Personalized ads on Facebook generate higher perceived privacy concerns than personalized ads on Instagram
2.5 Conceptual framework
The figure below depicts the conceptual framework of this study, which has been designed to test the hypotheses. The framework has been developed to analyze the effect of personalized advertisements on perceived privacy concerns in particular, on social media channels. In specific, the study examines whether there are differences between the perceived privacy concerns on Facebook and Instagram.
Figure 1: Conceptual framework
3. Research design & Methodology
To investigate the relationship between personalization of advertisements on social media channels and perceived privacy concerns, the hypotheses were tested empirically by collecting data. The first subsection describes the sample of this study followed by the research design, the procedure which includes an elaboration on the experimental survey and the measures used.
3.1 The sample
There were no restrictions regarding who can complete the survey as all age groups can use social media platforms. In total 204 people completed the survey of which 82 were male and 122 were female. Each participant was randomly allocated to one of the four treatments presented in the table below (Appendix II)
Figure 2: Experiment treatments
3.2 Research design
In order to identify the effect of personalized advertising on perceived privacy concerns, a vignette type of study was chosen. The study was conducted by embedding the experiment in an online survey in which participants were exposed to different treatments. A vignette study was appropriate for the purpose of this study because the judgments elicited by the participants are highly likely to be close to responses in a real
setting. Due to the limited scope of the study and the limited resources, it is impossible to personalize an ad for each participant based on the information they disclose on their social media accounts. Therefore, a vignette design was the most suitable method for this study. In total, 204 participants were randomly assigned to one of the four treatments in which they were exposed to either a personalized advertisement or a generic advertisement on one of the two social media platforms: Facebook or Instagram.
3.3 The procedure
The experiment was structured in four parts. In the first part of the experiment participants were instructed to complete questions regarding their demographics. The second part of the experiment was a filler task in which they were asked questions about their holiday destination preferences and random questions such as “ how many times per year do you go on holidays?” which are not relevant to the study. The filler task has been incorporated to mislead the participant into thinking that the experiment is about their holiday preferences. The reason why such a task was incorporated was to avoid revealing the true purpose of the study therefore preventing the “priming” effect. After completion of the filler task the participants were allocated to one out of the four treatments, which is the third part of the experiment. In each treatment the participants were presented with the same hypothetical situation in which they were asked to imagine that they are going on holiday to the Dutch Antilles. The hypothetical situation was followed by an existing KLM advertisement in either a Facebook or Instagram mobile application setting. Depending on the treatment they were in, participants got to see either an advertisement for tickets to the Dutch Antilles (personalized) or an advertisement for tickets to destinations in Europe. The reason why a mobile application setting was particularly exhibited was to minimize the differences as much
as possible between the two platforms therefore making the experiment more “clean”. After the advertisement was shown a series of questions followed regarding the relevance of the ad, whether they would click on it, whether they liked it and how much it intruded their privacy. The same set of questions were presented in all four treatments and no questions could be skipped. In the final part of the experiment the participants were asked about their social media usage.
3.4 Variables & measurements
The main dependent variable of the study is perceived privacy concern and the main independent variable is personalized advertising. Other independent variables included in the study were: social media platform (Facebook/Instagram) which was also the moderator in the model, likeability, click rate and relevance of the ad. All variables were measured with a 7-‐point Likert scale. According to Preibusch (2013) a Likert scale is a reliable way to measure privacy concerns.
4. Results
In this section, the hypotheses were tested by performing analyses in SPSS. The experiment was distributed online using Qualtrics for a duration of 4 days.
4.1 Preliminary analysis
The first step was to clean the data from inadequate results. In total the sample size amounted to N=204 participants with no missing values. The treatments were randomly distributed and equalized which resulted in the following allocation: 51 participants being allocated to treatment I, 53 to treatment II, 51 to treatment III and 49 to treatment IV which surpass the required number of 30 participants per treatment (Saunder & Lewis, 2012). The sample consisted of 40% male and 60% women, 63% of whom fall in the age group of 18 to 24 years (M=4.43). More than 65% indicate that they spend 0-‐2 hours a day on social media and more than 79% have both a Facebook and Instagram account. The data was grouped into conditions and the counter-‐indicative scale items were recoded. In addition to these steps, two dummy variables were created for the independent variables platform (Facebook/Instagram), which is also the moderator, and for personalization (yes/no).
4.2 Manipulation check
To examine whether there is a statistically significant difference between the means of the personalized and generic treatments, a one-‐way analysis of variance (ANOVA) was performed. The dependent variable used for the manipulation check was relevance of the ad. In other words, the treatments were compared to see if participants perceived advertisements in the personalized treatments as more relevant than the advertisements in the generic treatments. The graph below shows the mean relevance per condition.
Figure 3. The effect of personalization of ads on ad relevance
The graphical illustration portrays that the mean relevance of treatment I (M=5.14, SD=1.33) and treatment II (M=5.30, SD=1.25) is higher than the mean relevance in treatment III (M=3.08, SD=1.92) and treatment IV (M=3.82, SD=1.81). At first sight, these results suggest that overall the advertisements featured in the personalized treatments were perceived as more relevant than the advertisements in the generic treatments. The table below shows that there is a statistically significant difference between the treatments as determined by one-‐way ANOVA (F (3,200)= 23.00, p=. 00). Another noteworthy result is that in both the personalized and generic conditions, advertisements featured on Instagram generated higher means of relevance than advertisements on Facebook.
Table 2. One-‐way ANOVA of Relevance
SS DF MS F Sig.
Relevance 176.05 3 58.68 23.00 .00
Error 510.24 200 2.55
Total 686.29 203
Note: Significant at the p < 0.05 level
Additionally, a post-‐hoc Bonferroni test was performed to show which treatments differed from each other. In specific, the Bonferroni post-‐hoc test revealed that the personalized treatments (Facebook & Instagram) were not significantly different from each other (p=1.00). In contrast, the personalized Facebook treatment was significantly different than both generic treatments (p=.00). In line with the previous result, the personalized Instagram treatment was significantly than both generic treatments (p=.00). As for the generic Facebook treatment, the results show that it is not significantly different from the generic Instagram treatment (p=. 13). Additionally, a Contrast test was conducted to verify that the personalized ads significantly increased the perceived level of relevance compared to the generic ads t (173.33)=7,87, p= < 0.05 (two-‐tailed). The main finding from this analysis is that the personalized treatments and generic treatments are statistically different from each other, which implies that the personalized ads were perceived as more relevant than the generic ads. In sum, these results render the manipulation successful.
4.3 Formal test of the model
To determine whether personalized advertisements lead to greater perceived privacy concerns than generic advertisements, the hypotheses developed were tested. In this study, the hypotheses were tested at a significance level of p<0.05, meaning that results for the tested hypotheses with p-‐values higher than 0.05 were rejected. To test hypotheses 1 and 2, a Univariate ANOVA was performed in SPSS, which allows for the examination of the main effect and interaction effect. Consistent with the proposed theory, the results reveal that there is a significant main effect of personalization on perceived privacy concerns F (1,198)= 8.32, p=. 00 and the effect size is low (partial η 2=
.04). The results show that personalized ads generate higher perceived privacy concerns relative to generic ads therefore hypothesis 1 is supported. Furthermore, R2= .082 meaning that 8.2% of the total variation in privacy can be explained by personalization .In addition to testing the main effect, it was of interest to determine whether the moderator social media platform (Facebook & Instagram) interacted with the relation between the independent variable personalization and the dependent variable perceived privacy concerns. As illustrated in figure 4 below it was found that there is no significant interaction effect between personalization and social media platforms on perceived privacy concern after controlling for age (grouped in Quasi intervals) and gender F (1, 198) = .20, p=. 665. These results indicate that even though perceived privacy concerns were higher in the personalized Facebook condition (M= 5,02, SD= 1,44) relative to the personalized Instagram condition (M=4,25, SD=1,62), the difference between the means is not statistically significant. Intuitively, the effect of personalized ads on privacy concerns does depend not on the platform type. In sum, personalized advertisements on Facebook do not generate higher privacy concerns thus hypothesis 2 is not supported.
Interestingly, the results also revealed that the relation between age and perceived privacy concerns is significant (p=. 04) which implies that the higher the age group, the higher the levels of perceived privacy concern. To further investigate the relationship between age and perceived privacy concerns regardless of platform type and personalization, a Spearman’s correlation was conducted. The results revealed that the relationship between age and privacy is significant (rs=. 17, p=. 02). The findings of the study will be further elaborated on in the discussion and conclusion section.
Figure 4. Main effects and interaction effects