Threat or treat : personalized social network site advertising and consumer continuance intention

(1)

Threat or treat: Personalized social

network site advertising and consumer

continuance intention

Abel Somers (10073388) Master’s thesis Graduate School of Communication Master’s programme Communication Science Thesis supervisor: Dr. S. C. Boerman 30-6-2017 Word count: 7459

(2)

Abstract

The body of research on personalization in the context of social networking sites (SNS) is growing, but far from conclusive. This study contributes by studying the effects of advertisement personalization on user’s continuance intentions, and in how far this relation is influenced by a consumer cost-benefit analysis; the privacy calculus. Experimental results showed that moderate personalization positively affected continuance intentions, yet high ad personalization led to lower continuance intentions. Moreover, consumers were not found to consider these findings through a calculus of benefits and costs.

(3)

Introduction

As of this year, Facebook has passed an estimated 10.4 million users in the Netherlands alone. Daily usage of the social network site (SNS) is still increasing, indicating that 2017 will be another year of growth (Newcom, 2017). Networking sites such as Facebook have penetrated society, becoming indispensable in daily life. Allowing for a means of staying in touch with everyone around them, an estimated 89% of Dutch adolescents are users of this networking service (Newcom, 2017). Although social media have a range of advantages for their users, they are not the only ones profiting from the numbers addressed above. Information shared by users appears on the screens of friends, but is also funneled into the social networking site’s big data storage. This data is in turn used to deliver more personalized ads to users. Facebook’s revenue steadily increases each year, with marketers often heeding the temptation of SNS advertising (Smallwood, 2016). Targeting specific groups through personalized social media advertising has been an area of rapid growth, wherein practitioners outpace academic research (Hadija, Barnes & Hair, 2012; Louise, Kerr & Drennan, 2010).

The current study intends to reduce the lacking academic knowledge about the role of personalization on social network sites, specifically the medium Facebook. Since its launch in 2004, a multitude of studies have been done on topics related to the aspect of personalization (e.g., De Keyzer, Dens & Pelsmacher, 2015; Li, Liu & Jin, 2014; Tucker, 2014; Park, 2014). Findings on personalization include positive effects on click-through intention (De Keyzer et al., 2015), user continuance behavior (Li et al., 2014) and brand evaluation (Kalyanaraman & Sundar, 2006; Walrave, Poels, Antheunis, Van Den Broeck & Van Noort, 2016). Despite this, aspects of personalization remain unclear, with some of the recent and recurring ones being the optimal level of personalization (De Keyzer et al., 2015; Walrave et al., 2016), the mediating role of privacy calculus in context of different SNS (Dienlin & Metzger, 2016) and the influence of personalization on SNS user intentions (James, Warketin & Collignon, 2015). There has been scarce research on the boundaries of personalized advertisements. As a result, different personalization scales have been used throughout recent literature, thereby counteracting the aim of developing a clear personalization continuum (De Keyzer et al., 2015). This aspect is a focal point, with the current study intending to asses clearer boundaries of personalization.

(5)

5 Research has taken different stances on the determination of personalization levels, ranging from distinctiveness of personal information (White, Zahay, Thorbjørnsen & Shavitt, 2008) to customization expectations (Park, 2014) and perceived personalization based on personal information (Walrave et al., 2016). This study will adopt the view of personalization levels as used by Walrave et al. (2016), focusing on the consumer-perceived privacy sensitivity of personal information that SNS have access to. This is important because consumers have to actually experience an ad as personalized, preventing the occurrence of a mismatch between personalization and perceived personalization as stated in a previous study (De Keyzer et al., 2015). We use a low, moderate and high level of personalization to measure effects on user continuance intention. This relationship is expected to be positive, explained through the increased value and relevance that personalization offers to consumers participating in the SNS (Aguirre, Roggeveen, Grewal & Wetzels, 2016; Park, 2014). This would lead to a higher intention to continue using the SNS. We base personalization levels on personal data categories such as name, address, birth date and relationship updates.

An explanation for consumers’ usage continuance of SNS, despite resistance to personalized advertisements, is the mediating role of consumers’ privacy calculus (Dienlin & Metzger, 2016). Founded in economics, it is now being applied to communication science research on SNS (Krasnova, Spiekermann, Koroleva & Hildebrand, 2010; McKnight, Langton & Knipp, 2010). Privacy calculus theory proposes that consumers weigh the perceived benefits and costs that are attached to certain action. When related to SNS as Facebook, one can view a user as a homo economicus, actively weighing gains such as social interaction and increased ease of use, to the possible privacy losses that are involved in sharing personal information with Facebook (and other organizations indirectly). Due to the only recent application of privacy calculus to SNS, much is yet to be researched (Dienlin & Metzger, 2016). One aspect of uncovered research is the consequential intentions and actual behavior that stem from consumers’ privacy calculus (James, Warketin & Collignon, 2015). User continuance intention is defined as the degree to which consumers intend to continue usage of SNS in this study’s context.

The central aim of this study will be to clarify the role of different levels of personalized social network advertising in predicting user continuance intention, mediated by the privacy calculus of consumers. This study further contributes by providing a more thorough understanding of the way consumers perceive personalization, leading to practical

(6)

6 guidelines of personalization boundaries for both consumers and organizations. To guide this study, the following research question will be central:

To what degree do different levels of personalization in Facebook advertisements impact consumers’ user continuance intention of the social network site, and is this mediated by consumer privacy calculus?

Theoretical framework

Consumer responses to online advertisement personalization

Although a broad scale of personalization definitions can be found in literature, some aspects recur; it is an effort by the sender to create distinct messages that seem tailored to an individual’s personal demands, needs and characteristics. The message thereby refers to the recipient’s self without actually changing the message content (Maslowska, Smit & Van Den Putten, 2011, 2016; White et al., 2008). Importantly, it must also be perceived as such by the recipient, as a mismatch does not grant any actual personalization effects onto the targeted individual (De Keyzer et al., 2015). A distinction is made between personalization and a related form of customer-oriented marketing, customization (Aguirre, Mahr, Grewal, De Ruyter & Wetzels, 2015; Arora et al., 2008). While personalization is an effort by the sender, customization is dependent on preferences as proactively poised by the recipient, after which the sender actually customizes the message.

Some behavioral consequences of personalization remain insufficiently studied. One area in need of empirical investigation concerns predictors of social network site usage continuance intentions (Ku, Chen & Zhang, 2013), in particular the effect that personalized advertisements have on social network site usage intentions (James et al., 2015). Based on the elaboration likelihood model (ELM), this relation can be predicted and explicated (Ho & Bodoff, 2014). Personalized messages may lead to consumer perceived personal relevance by incorporating information of the recipient’s self, thereby accommodating their demands or needs (Aguirre et al., 2015; Dijkstra, 2008; O’Donnell & Cramer, 2015). Individuals are sensitive to self-relevant and self-referencing information. Personalization cues that match with the self kindle cognitive attention, leading to elaboration on the message information. The ELM includes this elaborate processing as central processing, which is accompanied by high cognitive effort. When an individual is motivated and able to attentively indulge the personally relevant persuasive message, impacting relevant attitudes and consequent behavior becomes easier and more persistent for marketers (Maslowska et al., 2011; De Keyzer et al.,

(7)

7 2015). Mind that the opposite may occur as well; if negative thoughts outweigh positive ones, personalized advertisement effects may weaken through critical elaboration (Hawkins et al., 2008; Kelly, Kerr & Drennan, 2010). When an individual has no motivation or ability to dedicate cognitive resources to a message, it is peripherally processed. Although superficial aspects of a message such as its source may have some effect, peripheral processing is less effective at rendering persuasive effects onto attitudes and consequent behavior (Maslowska et al., 2011).

Previous studies have shown that personalized advertisements can positively affect persuasive effects in different message settings (Walrave et al., 2016), by increasing consumer engagement as a consequence of attentional processing, increasing customer loyalty, satisfaction and retention (Basak & Calisir, 2015; Chellappa & Sin, 2005; Maslowska et al., 2016), a more positive attitude (Kalyanamaran et al., 2006), better memorization and liking of personalized messages (De Keyzer et al., 2015; Li, 2016) and consumer enjoyment due to increased service and message relevance (Aguirre et al., 2015; Hawkins et al., 2008).

However, negative effects may occur as well; consumers may perceive the message as irritating, intrusive or as interfering with their goal pursuit (Van Doorn & Hoekstra, 2013). The major disadvantage of ad personalization for consumers is that it may lead to privacy concerns through feelings of discomfort and intrusiveness (Tucker, 2014). When a message is perceived as too personal, it may lead to reactance (White et al., 2008), mainly caused by privacy concerns such as perceived vulnerability and risk (Aguirre et al., 2015; Li, 2012).

Findings show both positive and negative effects, but personalization overall seems beneficial for the consumer (Walrave et al., 2016). Mixed findings may be explained by the difference between the measurement of perceived personalization by consumers, and personalization as seen by researchers. Recent studies argue that perception and actual personalization ought to be treated as separate constructs (Li, 2016; Maslowska et al., 2016). Therefore, the current study will incorporate the measurement of perceived personalization. Literature has a recurring indication that higher levels of personalized content have positive effects, yet are also more vulnerable to negative effects (De Keyzer et al., 2015; Van Doorn & Hoekstra, 2013). Low degrees of personalization had a less positive effect, but no negative effect onto engagement and usage continuance (Aguirre, 2016; Park, 2014). Bleier and Eisenbeiss (2015) also found relevantly personalized ads to have a positive effect, as long as consumers had trust in the agent.

(8)

8 Altogether, more personalization of an advertisement message will increase the personal relevance for the recipient through self-referencing information. The central processing of the information then positively affects social network site usage continuance intention. The more personalized an advertisement is, the higher the chance that the user will positively evaluate it. Therefore, the following hypothesis is formulated (depicted in Figure 1):

H1: A high level of advertisement personalization leads to a more positive usage continuance intention by consumers than a moderate level of advertisement personalization, and a low level of advertisement personalization will lead to the least positive usage continuance intention.

Privacy calculus

Besides personalization, other predictors of continuance (intention) have found to be the privacy concerns, as well as usefulness and SNS enjoyment (Krasnova & Veltri, 2010; McKnight et al., 2010). Other research has found personalization to affect experienced gratifications (Ku et al., 2013), switching costs and satisfaction (Park, 2014). These factors consisting of benefits and costs associated with personalized advertisements on SNS can be related to the privacy calculus concept, which will now be discussed.

In the last decade, opposing factors involved in privacy decision making were studied under the name of privacy calculus (PC) in SNS (e.g. James et al., 2015; Krasnova & Veltri, 2010; Kehr, Wentzel & Mayer, 2013; Zhu, Ou, Van den Heuvel & Liu, 2016). The idea behind the privacy calculus theory is that behavioral action is determined by an active, rational consideration of the benefits and costs. If benefits outweigh the privacy-related costs associated with the release of personal data, individuals will respond accordingly and engage in behavioral action such as self-disclosure (Chellappa & Sin, 2005). On the other hand, one will refrain from action if costs weigh heavier in an individual’s calculus (Krasnova & Veltri, 2010). Although, in practice this straight-forward relationship seems more difficult to assess due to the privacy paradox phenomenon, which describes how consumers disclose information while, paradoxically, stating to not feel like doing so due to high perceived privacy risks (Dienlin & Metzger, 2016).

Receiving personalized advertisements on a social network site can lead to an individual experiencing benefits such as higher satisfaction (Chellappa & Sin, 2005; Maslowska, 2016; Park, 2014) and increased enjoyment (Aguirre et al., 2015; Hawkins et al.,

(9)

9 2008). Other benefits are personalization itself, financial rewards and social adjustment (Smith, Dinev & Xu, 2011). Additionally, Krasnova and Veltri (2010) found enjoyment, self-representation and relationship maintenance to be benefits, while others found entertainment to be a benefit of SNS usage (Dienlin & Metzger, 2016). Other, indirect benefits include increased social capital, social support, identity management and personal relevance (De Keyzer et al., 2015). Based on these findings, increased personalization is expected to enlarge the perceived benefits of an individual’s privacy calculus:

H2a: A high level of advertisement personalization leads to more perceived benefits by consumers than a moderate level of advertisement personalization, and a low level of advertisement personalization will lead to the least perceived benefits.

Besides being influenced by personalization, perceived usefulness, satisfaction, enjoyment and gratifications were also found to be predictors of SNS user continuance intentions (Basak & Calisir, 2015; Ku et al., 2013; McKnight et al., 2010; Sun, Liu, Peng, Dong & Barnes, 2013). As individuals receive more relevant and useful information, the benefits of continuous use should increase. These effects have been found in the online context of both Facebook and other SNSs (Mcknight et al., 2010; Yin, Cheng & Zhu, 2011). Other benefits mentioned above, such as relationship maintenance and self-representation are also expected to increase usage continuance intentions by making SNS usage more valuable. Based on the above, the following hypothesis is formulated:

H2b: Perceived benefits by consumers will positively influence usage continuance intention.

Lastly, consumer perceived benefits are also expected to have a mediating role. As personalization of SNS advertisements increases self-relevant information, consumers are more likely to elaborate on the information through central processing. While this is expected to increase continuance intentions, additional satisfaction because of the benefits of personalization can positively bolster this effect (Li, Liu & Jin, 2014). Therefore, the last privacy calculus’ related benefit hypothesis is as follows:

H2c: The positive effect of advertisement personalization on usage continuance intention will be mediated by perceived benefits, such that higher perceived benefits lead to more positive usage continuance intentions.

Increased personalization of advertisements on SNS can also lead to privacy concerns (Kehr et al., 2013). As privacy itself is a concept that is difficult to measure, empirical

(10)

10 research focuses on the measurement of privacy concerns, based on factors such as concerns about data collection, errors, unauthorized use of personal information and use of personal information by secondary sources ((Davies, 1997; Smith et al., 2011). Perceived privacy risks as a preceding concept has found to be another reliable determinant of SNS usage, defined as the degree to which a person believes that SNS usage may lead to negative consequences for privacy (Ernst, 2014).

As effortful processing of a message increases through personalization, this may increase the amount of negative thoughts towards the message (Maslowska et al., 2016). Exposure to a highly personalized message increases the chances that it is seen as illegitimate and is critically assessed, leading to higher perceived costs and skepticism towards the personalized message (White et al., 2008). Costs associated with SNS privacy calculus are found to be the perceived risk of informational disclosure, likelihood and damage of privacy violations, sensitivity of personal information, loss of disclosure control and perceived SNS switching cost (Dienlin & Metzger, 2016; Krasnova & Veltri, 2010; McKnight et al., 2010; Park, 2014). Based on the above, the following hypothesis is formulated:

H3a: A high level of advertisement personalization leads to more perceived costs by consumers than a moderate level of advertisement personalization, and a low level of advertisement personalization will lead to the least perceived costs.

Privacy concerns have also been found to be a negative predictor of actual user continuance behavior in past research on SNS (Chen, 2013; Zhou & Li, 2014), as well as in the context of online services (Pavlou, Liang & Xue, 2006) and e-commerce (Ku et al., 2013). As such, negative thoughts and risks associated with the use of a medium are expected to lower an individuals’ motivation to continue using the SNS, leading to the following hypothesis:

H3b: Perceived costs by consumers will negatively influence usage continuance intention.

Furthermore, the main relationship between advertisement personalization and usage continuance intention is expected to be mediated by the perceived costs of a consumer. Personalization may increase an individual’s perceived benefits, but may also be found worrisome and lead to negative responses (White et al., 2008). Previous research confirms the consequential expectation, that upped costs associated with the reception of a more personalized ad message lower the intention of said consumer to continue using the SNS (Ku et al., 2013).

(11)

11

H3c: The positive effect of advertisement personalization on usage continuance intention will be mediated by perceived costs, such that higher perceived costs lead to more negative usage continuance intentions.

Figure 1. Proposed mediation model, in which higher advertisement personalization increases

user continuance intentions, mediated by privacy calculus’ perceived benefits and costs.

Methods

Sample and design

We conducted an online experiment with a single-factor experimental design among Dutch (young) adults between the age of 18 and 60, wherein the ordinal level factor Perceived advertisement personalization was treated as a between-subject variable (3 levels: High, moderate, and low perceived personalization). The age group includes people that are among the most active users of social media, and more relevantly the medium Facebook (CBS, 2015; De Keyzer et al., 2015). Participants below 18 years old were not targeted because adolescents between 10 and 17 years of age have different disclosure patterns of personal data for marketing purposes (Walrave et al., 2016). Participants were approached through

(12)

12 convenience sampling. A recruitment message was spread among multiple private as well as public Facebook groups and pages. These groups included members that fit the age and Facebook usage criteria. In total, 224 respondents participated in the questionnaire. Twenty-nine respondents did not complete it and were excluded, leaving 195 respondents that completed the questionnaire. In the final sample, only those that were 18 years or older, owned a Facebook account and had a Dutch language proficiency were included (N = 194). The sample comprised of 32.5 percent men and 67.5 percent women. The participant’s ages ranged from 18 to 55 years, with a mean of 23.69 years (SD = 3.46). The average educational level was an academic bachelor’s degree, and 68.5% had a volitional education level or higher.

Stimulus materials

Perceived advertisement personalization.

Similar to Walrave et al. (2016), a pretest was conducted to create three different conditions for the variable Perceived advertisement personalization. Respondents were asked to rate fourteen different types of information that can be found on a Facebook profile, in terms of how personalized they would experience the usage of these information types in an advertisement (see appendix A1). Four related seven-point Likert scales were used for measurement. These were ‘very unpleasant (1)’ to ‘very pleasant (7)’, ‘very useless (1)’ to ‘very useful (7)’, ‘very unacceptable (1)’ to ‘very acceptable (7)’ and ‘very impersonal (1)’ to ‘very personal (7)’ (reverse coded). Calculated means as a result of the pretest (N = 18) led to the selection of the item Gender (M = 4.74, SD = 0.89) as a low personalization condition, the item Date of birth as a moderate personalization condition (M = 3.29, SD = 1.17), and the item Cell phone number (M = 1.58, SD = 0.57) as a high personalization condition.

The three conditions were operationalized by editing a Facebook news feed screenshot. An advertisement about sunglasses could be found on the right side of the Facebook page, exactly at the natural location of a domain advertisement. For the Gender condition, an advertisement on the Facebook page included information based on the answer of the question about gender. If man was selected, the advertisement included the text ‘As a man, you get 30% discount’ (see appendix B1) and if woman was selected the text ‘As a woman, you get 30% discount’ (see appendix B2). The Date of birth condition had a similar domain advertisement that included the text lines ‘birthday discount’ and ‘Congrats! Get your 30% discount now’ below the picture of sunglasses (see appendix B3). The domain

(13)

13 advertisement of the third condition Cell phone number contained the text line ‘Is 06-20334187 still your mobile number?’ (see appendix B4).

Measures

User continuance intention.

The dependent variable User continuance intention was measured using the slightly adapted seven-point Likert scales as used by McKnight et al. (2010), scaling from scores between ‘completely disagree (1)’ to ‘completely agree (7)’. A principal component analysis (PCA) was conducted on the three items with oblique rotation (direct oblimin). One component was extracted (Eigenvalue = 2.53, variance explained = 84.42%) with Cronbach’s α = .90, showing high scale reliability. The items included were “In the near future, I intend to continue using Facebook”, “I intend to continue using Facebook” and “I predict that I will continue using Facebook”. The variable User continuance intention was made by creating a mean score on these three items (M = 5.62, SD = 1.38).

Perceived benefits.

The mediator Perceived benefits was measured using scales from recent research (Dienlin & Metzger, 2016). A seven-point Likert-scale for the 10 items was used in this study instead of a five-point Likert-scale to get more refined answer options. Scale ranged from ‘completely

agree (1)’ to ‘completely disagree (7)’. Four items were reverse coded to prevent

acquiescence bias. An oblique rotation (direct oblimin) PCA was conducted on the 10 items, resulting in the extraction of three components (Eigenvalues = 2.98, 1.82 and 1.12), together explaining 59.20% of the variance. Analysis of the pattern matrix indicated that the items that clustered on component 1 represented controlling relationships with others through Facebook, while component 2 represented understanding and expressing emotions and beliefs, whereas component 3 clustered around connecting with others. The component correlation matrix indicated that all three components were as expected related, being aspects of perceived benefits.

Based on the previous, no components or items were excluded from the new variable Perceived benefits (M = 4.26, SD = 0.89), created by calculating a mean score for all 10 items together. This variable had an overall good scale reliability (Cronbach’s α = .72). Individual item reliability was good, and exclusion of any item would not increase the overall reliability. Perceived benefits items included “The use of Facebook helps me come in contact with others that share my views and interests, and maintain this contact”, “The use of Facebook helps me

(14)

14 feel more connected to others” and “The use of Facebook does not help me make new social or business contacts” (see appendix C1).

Perceived costs.

Perceived cost items focused on privacy concerns, the generally adopted measurement of SNS costs in literature (Dienlin & Metzger, 2016). Two additional items were based on McKnight et al. (2010), measuring the same construct privacy concern to increase scale reliability. All items were measured on a seven-point Likert scale (1= ‘completely disagree’, 7 = ‘completely

agree’). Perceived cost items include: “I feel concerned about my privacy online”, “I do not

worry about my privacy as a result of using Facebook”, “The danger to people’s privacy when they use social networking sites like Facebook has been overblown”. Appendix C2 provides an overview of all items. The average score on the total of six items was taken to create a mean score of Perceived cost (M = 4.71, SD = 1.06). Two items were reverse coded. A principal component analysis (PCA) with oblique rotation (direct oblimin) was conducted on

the six items, resulting in the extraction of one component that all items clustered on (Eigenvalue = 3.14, variance explained = 52.31%). Reliability analysis indicated that the

Perceived costs scale was good, with Cronbach’s α = .81.

Manipulation check.

At the end of the questionnaire, respondents were asked if they recalled seeing an advertisement (‘Yes’ or ‘No’), followed by a question on how personal the experimental stimulus was perceived by respondents (‘very personal information (1)’ to ‘very impersonal

information (7)’) (see appendix A2). Lastly, we checked for recognition of the manipulation

by asking respondents what type of advertisement personalization they were exposed to, which could be answered with ‘Gender’, ‘Cell phone number’, ‘Date of birth’ or ‘Did not see,

don’t know’.

The three types of information used in the experimental conditions were asked for a second time at the end of the questionnaire as a manipulation check. Respondents indicated on the same questions as the pretest how personalized they would perceive an advertisement including information about gender, birth date and cell phone number.

Control variables.

This study included control variables to ensure that the experimental manipulation was responsible for the found effects. Participants were asked for their age (M = 23.69, SD = 3.46) and gender (32.25% male). Educational level was asked, with the average educational level

(15)

15 being an academic bachelor’s degree. Participants were also asked whether they had a Facebook account (99.49%). Daily Facebook usage was measured with seven options ranging from ‘less than 15 minutes’ to ‘more than 6 hours’. The average time spent per day was 30 to 60 minutes.

Results

Manipulation check

The manipulation check revealed that overall recall of the ad (n = 183) and correct recognition of the ad information personalization type was low (n = 87) with 44.85%. Based on this, all consequent analyses were performed once with an exclusion of respondents that did not pass the manipulation check (gender n = 41, date of birth n = 23, cell phone number n = 23), and once with all respondents (gender n = 66, date of birth n = 66, cell phone number n = 62). Analyses with both respondent groups were done to establish whether respondent stimulus recognition and recall had an effect on the findings. Analyses with the respondents that passed the manipulation check will be discussed first, as these findings are based on the more reliable data. These are followed by brief observations on the results including all respondents if relevant findings were done compared to the subsample results. An elaboration on manipulation check information can be found in appendix D.

Respondents were also asked in the questionnaire manipulation check to rate the experimental condition’s information types, similar to the pretest. The results show a similar order and Likert scale score as the pretest (see appendix A2).

Randomization

The three experimental groups in the subsample did not significantly differ regarding age

F (14, 72) = 0.772, p = .695; education, χ² (14) = 11.79, p = .624; gender, χ² (2) = 0.30, p =

.860 and Facebook use, χ² (8) = 14.91, p = .179. Similar non-significant effects were found for the total sample, including respondents that did not pass the manipulation check.

Effects on user continuance intention

A multi-variate analysis of variance (MANOVA) was conducted to test H1, establishing whether the three experimental conditions (gender, date of birth and cell phone number) had significant effects on the dependent variable User continuance intention. Type of information personalization was inserted as independent variable, and user continuance intention, perceived benefits and perceived costs as dependent variables. Results showed a marginally

(16)

16 significant main effect of information personalization type on perceived benefits, costs and user continuance intention using Pillai’s Trace, V = .128, F (3, 82) = 1.89, p = .086. However, separate univariate ANOVAs revealed that there was no significant effect of different types of advertisement personalization on respondent’ perceived benefits F (2, 84) = 0.91, p = .407

and costs F (2, 84) = 1.45, p = .240, only a significant effect on user continuance intention

F (2, 84) = 3.28, p = .043 (see Table 1).

Table 1

Effect of advertisement personalization type on perceived benefits, costs and user continuance intention

Gender Date of birth Cell phone number

Perceived benefits 4.36 (1.01)a 4.12 (0.93)a 4.49 (0.81)a Perceived costs 4.62 (1.10)a 5.04 (1.02)a 5.00 (1.17)a User continuance intention 5.69 (1.17)ab 6.22 (0.77)a 5.30 (1.60)b

Note. Mean scores with standard deviations in parentheses.

a,b

Means with different superscripts (within the same row) significantly differ from each other with p < .05.

Bonferroni corrected post-hoc comparisons showed that respondents exposed to the date of birth condition had a significantly higher continuance intention compared to respondents in the cell phone number condition (Mdifference = 0.91, p = .038). There was no

significant difference found between the effects of the gender and date of birth conditions (Mdifference = -0.53, p = .301) or the gender and cell phone number conditions (Mdifference = 0.39,

p = .677) on user continuance intention. Similarly, none of the perceived personalization

conditions had a significant effect onto the mediators perceived benefits and costs. A MANOVA for the total sample (N = 194) showed no significant effects (p > .411).

Based on these results, H1 was not supported. The post-hoc procedure showed a significant difference between exposure to the date of birth and cell phone number conditions. The expected effect of a higher continuance intention when exposed to an advertisement containing a birth date compared to gender information, was not found. Respondents exposed to the date of birth condition did have a higher continuance intention than those in the cell phone number condition. This is -interestingly- contrary to the hypothesized relation, in which

(17)

17 higher perceived personalization of an ad would lead to a higher user continuance intention. The next section will further discuss the main effect and the role of the hypothesized mediators.

A parallel mediation model was tested using Model 4 of the PROCESS macro in SPSS (Hayes, 2013). This method uses 5,000 bootstrap samples to create a reliable confidence interval, and calculates regressions for the total and direct effects of perceived advertisement personalization onto user continuance intention via mediators perceived benefits and perceived costs. A dummy variable was created for each condition to test for differences between the three experimental conditions. Three separate mediation analyses were then conducted (see Figure 2, Table 2), wherein the independent variable and covariate inserted differed to cover all possible comparisons. A third experimental condition was always left out of the mediation analysis as reference category.

Figure 2. Mediation model, in which higher advertisement personalization increases user

(18)

18 Results of the mediation analyses showed, similar to the MANOVA, that there was a significant direct effect (controlling for mediators) of exposure to the date of birth condition

compared to the cell phone number condition on user continuance intention (bc’ = 0.87, p = .016). Exposure did not significantly differ when comparing the gender condition to the

cell phone number condition (bc’ = 0.27, p = .390), but gender compared to date of birth (bc’ =

-0.60, p = .060) was found to be marginally significant in the mediation analysis, the gender ad thus leading to a lower user continuance intention than the date of birth advertisement personalization. These findings partially support H1.

A correlation analysis was additionally performed to ensure mediator independence. This was the case, with a very weak correlation that was non-significant between perceived benefits and perceived costs for the subsample that passed the manipulation check (N = 87),

r = .10, p = .356. Similar results were found for the total sample (N = 194), r = .10, p = .168.

Hypotheses H2a, H2b and H2c on the mediating role of perceived benefits were tested through the mediation analysis. Findings showed that H2a was not supported, as the moderate advertisement personalization condition date of birth did not lead to a significant decrease in perceived benefits compared to cell phone number (ba = -0.37, p = .191), or increase when compared to the gender condition (ba = 0.24, p = .331). Perceived benefits were also not significantly lower for gender compared to the cell phone number advertisement personalization (ba = -0.13, p = .608).

Hypothesis 2b considering the positive effect of perceived benefits onto user continuance intention was not supported with no significant effect (bb = -0.14, p = .323); higher perceived benefits did not lead to a higher continuance intention.

Table 2

Mediation: Effect of personalization level on user continuance intention via perceived costs and benefits

Group (reference) abenefits acosts bbenefits bcosts c c’ Indirect effect

[95% BCAB] benefits costs Gender (Cell phone number) -.13

(.24) -.38 (.29) -.14 (.14) -.26 (.12)* .39 (.32) .27 (.31) .02 (.05) [-.04; .16] .10 (.09) [-.02; .34]

(19)

19 Gender (Date of birth) .24

(.24) -.42 (.29) -.53 (.32) -.60 (.32)** -.03 (.06) [-.27; .03] .11 (.09) [-.01; .35] Date of birth (Cell phone

number) -.37 (.28) .04 (.32) .91 (.36)* .87 (.35)* .05 (.07) [-.03; .30] -.01 (.09) [-.21; .16]

Note. A, b, c and c’ are unstandardized b-coefficients. Standard error between brackets. BCAB = Bias corrected and accelerated confidence interval. * p < .05, ** p = .06.

Determining the mediating overall role of perceived benefits, H2c, results showed that the total effect of the gender condition on user continuance intention was non-significant when compared to cell phone number (bc = 0.39, p = .226) and date of birth (bc = -0.53, p = .100). The total effect of the condition date of birth on user continuance intention was found to be significantly higher in comparison to the cell phone number condition (bc = 0.91, p =

.013). Controlling for mediators perceived benefits and perceived costs, the direct effect of the gender condition on user continuance intention remained non-significant when compared to cell phone number (bc’ = 0.27, p = .389), but was marginally significantly different compared

to date of birth (bc’ = -0.60, p = .060). Similar results were found for date of birth compared to cell phone number (bc’ = 0.87, p = .016). Bootstrapping showed no significant indirect effect

of exposure to the gender advertisement onto user continuance intentions via perceived benefits (indirect effect = 0.02, boot SE = 0.05, 95% BCAB [-.04; .16]) compared to the cell phone number condition. No significant effects were found either for comparison to the date of birth condition with mediator perceived benefits (indirect effect = -0.03, boot SE = 0.06, 95% BCAB [-.27; .03]). No significant indirect effect of exposure to the date of birth condition onto user continuance intentions via perceived benefits (indirect effect = 0.03, boot SE = 0.07, 95% BCAB [-.09; .19]) was found compared to the cell phone number condition. Concluding, although the gender ad led to a significantly higher user continuance intention than the cell phone number personalization condition, and the cell phone number condition led to a significantly higher user continuance intention compared to the date of birth condition, no significant mediating effects were found. Perceived benefits did not increase regardless of the advertisement personalization type. The perceived benefits respondents envisioned then did not lead to a higher user continuance intention. H2c was thus not supported.

The total sample (N = 194) mediation analysis showed a significant indirect effect of the date of birth condition compared to the cell phone number condition onto user continuance intentions via perceived benefits (indirect effect = 0.11, boot SE = 0.06, 95%

(20)

20 BCAB [.02; .27]), a positive effect of high personalization on perceived benefits (ba = -0.33, p = .035), and an unexpected significant negative effect of perceived benefits on user continuance intention (bb = -0.32, p = .004). The significant effect of date of birth compared to cell phone number onto user continuance intentions disappeared when controlling for perceived benefits and costs (bc’ = 0.05, p = .845).

To test H3, the same mediation analyses were used. Results showed no support for H3a. The moderate advertisement condition date of birth did not lead to a significant decrease in perceived costs compared to cell phone number (ba = 0.04, p = .911), or increase when

compared to the gender condition (ba = 0.42, p = .147). Perceived costs were not significantly

decreased by a gender advertisement when compared to the cell phone number ad personalization (ba = -0.38, p = .191).

Hypothesis 3b considering the negative effect of perceived costs onto user continuance intention was supported (ba = -0.26, p = .033). When respondents had a higher score on the

perceived costs of Facebook use, they had a lower usage continuance intention.

The mediating role of perceived costs as proposed by H3c was not supported. The gender advertisement compared to the cell phone number advertisement condition did not

significantly increase user continuance intentions via mediator perceived costs (indirect effect = 0.10, boot SE = 0.09, 95% BCAB [-.02; .34]). No significant effects were

found either for gender compared to date of birth (indirect effect = 0.11, boot SE = 0.09, 95% BCAB [-.01; .35]). There was also no significant effect found between exposure to the date of birth and cell phone condition onto continuance intentions via mediator perceived costs (indirect effect = -0.01, boot SE = 0.09, 95% BCAB [-.21; .16]). In sum, results show that regardless of the advertisement personalization, it did not lead to significantly higher perceived costs, thereby leading to lowered user continuance intentions. Total respondent sample analysis found similar results.

Conclusion and discussion

As social network site participation keeps growing, social media advertising does as well. This study examines whether exposure to personalized advertisements affect Facebook usage continuance intentions, and studies the role of consumer’s privacy calculus as a mediator.

(21)

21 Findings on the main relationship between ad personalization and continuance intentions showed -as hypothesized- that exposure to moderate personalization of an advertisement led to a higher usage continuance intention than exposure to a lowly personalized ad. This is in line with the Elaboration Likelihood Model, as this study seems to indicate that higher relevance of a personalized message motivates effortful processing, positively affecting attitudes and thereby intentions (Oenema, Tan & Brug, 2005). Yet, the moderately personalized birth date advertisement led to a higher user continuance intention than the cell phone condition, contrary to the hypothesized relationship. Explainable through the elaboration likelihood model, deeper and more thorough processing is accompanied by more critical evaluation (Hawkins et al., 2008). This is what seems to have been the case in this study, with the highly personalized advertisement leading to a negative effect on continuance intentions compared to moderate advertisement personalization. A second explanation may be psychological reactance; when an individual perceives a message as too personalized, this may be experienced as an invasion of privacy (White et al., 2008). The sense of constrained personal freedom then triggers the individual to resolve this situation and reestablish freedom (Walrave et al., 2016). Concluding, the findings show that advertisement personalization does indeed affect continuance intentions of Facebook users. Although a higher continuance intention was only found when comparing moderate to low levels of personalization, the decrease in user continuance intentions for a high level of personalization can be related to past research. Therein, high personalization was found to have adverse effects onto click-through intentions and purchase intentions (Aguirre et al., 2015; Van Doorn & Hoekstra, 2013; White et al., 2008).

We also tested whether the effect of personalization on continuance intentions was mediated by the privacy calculus. We found no evidence that perceived personalization influences user continuance intentions through consumer’s privacy calculus. Hence, it provides no explanation in this study for the found negative effect of high ad personalization, nor the positive effect on continuance intentions of moderate personalization compared to low personalization. The non-significant effects may be explained through an absence of active calculus-like cogitations due to a habitual pattern, as suggested by Mcknight et al. (2010). As respondents’ use of Facebook on a daily basis averaged between 30 and 60 minutes, it is safe to say that usage is subject to habit formation. Additionally, the respondent subsample that passed the manipulation check showed overall high intention to continue using Facebook regardless of the amount of personalization they were exposed to (M = 5.73, SD = 1.25). This

(22)

22 may be a confirmation that a conscious privacy calculus is not something actually executed by consumers in this study, and that perceived costs associated with Facebook usage have moved to the background as suggested (Jung, McKnight, Jung & Lankton, 2011). Like Wood and Neal (2007) describe, a formed habit has conservative characteristics; increased usage of personal information in Facebook ads may not quickly influence habitual performance. While initial Facebook use may have been influenced by a privacy calculus, it could currently be the case that associated costs are taken for granted, or at least do not have a strong enough impact to influence continuance intentions. Hallam and Zanella (2017) show that privacy concerns relate to psychologically distant intentions and are therefore marginalized, bolstering this idea of a habitual overrule. Direct effect of this habitual behavior on user continuance intentions was found significant in the related SNS context of Twitter (Barnes & Böhringer, 2011).

Lastly, although we found no evidence of a privacy calculus, perceived costs did have a significant negative effect on user continuance intentions. This effect of perceived privacy concerns, the measurement of costs in this study, has been found in past research on a multitude of related topics including SNS, blogs and location-based social networks (Zhou & Li, 2014). We thereby confirm the relation and extend it to the context of Facebook.

This study has found SNS advertisement personalization to be subject to a tipping point, leading to a lower user continuance intention at high levels of personalization. While previous research has found such an effect in related context, this study is the first to ascertain it in the context of SNS. Thus, it offers practical implications. First and foremost, social networking platforms, such as Facebook, must think carefully about adding more personalization features to their advertisements, to ensure that ad profits do not interfere with actual user continuance. As this study finds, if Facebook would choose to add the option for advertisers to personalize ads based on cell phone number, they are at risk of deterring users, perceiving this kind of advertisements as excessively personal. A second, related implication comes forth from the negative effect of perceived costs on consumer continuance intentions. Although gathering data from social network site users may have positive effects such as increased usefulness and enjoyment (McKnight et al., 2010), it is important for Facebook to reduce its users’ privacy concerns to avert stopped usage of the medium. Other studies have provided strategies to mitigate these costs. Practical advice is to further increase privacy protection policies and inform users of its existence, as well as to improve the trust of users by providing safety cues (Ku et al., 2013; Zhou & Li, 2014).

(23)

23 This study is also subject to limitations. Although the manipulated Facebook pages are approximately identical to an authentic Facebook page, the content of the advertisements could have been made more personalized, as only a low amount of respondents (44.85%) passed the manipulation check. This unsuccessful manipulation is likely responsible for the different findings of the full sample analysis. The root cause of the failed manipulation may be lacking self-referencing information. The information included in the ads was perceived as personal (see appendix A2), but did not include actual individualized information. The high personalization condition, for example, contained a random cell phone number instead of the specific number of the participating respondent. Therefore, although the information type was perceived as personal, the actual ad itself did not to such a degree that it led to recall and recognition by respondents as expected from self-referencing information (Leshikar, Duas & Duarte, 2015). Similarly, Maslowska et al. (2016) found self-reference in a personalization context to be a significant predictor of attention, which precedes increased recall and recognition. It is anticipated that research including respondent-specific personal information would increase statistical power via the sample size with higher recall and recognition, possibly finding stronger effects of the different personalization conditions onto continuance intentions, as well as the privacy calculus.

A second limitation is the measurement of intentions as opposed to actual behavior. Although intentions have been found strong predictors of behavior, a gap between the two oftentimes exists (Sheeran, 2002). As Smith et al. (2011) state, future costs and benefits such as those of the privacy calculus have a tendency to be discounted, in so that there is a discrepancy between what consumers say they would do, and what their actual behavior would be. Therefore, if user continuance itself could have been measured in this study, there is a fair possibility of finding different and possibly significant calculus effects.

As this study’s field of research is far from saturated, it leaves much opportunity for future research. One area in need of more examination, are the effects of additional types of consumer perceived personalized information. As discussed earlier, there is a lacking consistency in the approach of the concept personalization. By adopting this study to look at other types of personal information, research may work towards the development of a personalization continuum with corresponding effects. This is especially relevant as previous studies -as well as this study- have located tipping points after which negative personalization effects occur: Limits include the use of personal identification and online transaction data (Doorn & Hoekstra, 2013), five items of personal information such as name and phone

(24)

24 number (White et al., 2008) and combined demographic information types such as age, gender and location (Aguirre et al., 2015). It seems therefore valuable for the field of study to take a unified approach at not only studying personalization in an online context, but also its effects on continuance intentions and behavior.

Finally, as no mediating effect of a consumer privacy calculus was found, a

recommendation for future research is to further study the role it plays. While there may be explanations for the lacking significant effects of privacy calculus in this study, past research did find significant positive and negative effects of personalization (De Keyzer et al, 2015; White et al., 2008), as well as their effect on behavioral action (Chen, 2013; Mcknight et al, 2010). A suggestion would be to alter the measurement of privacy calculus’ costs and benefits. As elaborated before, once Facebook usage has become habitual, it may no longer lead to an active calculus. Future personalization research should attempt to measure the privacy calculus in a context of actual behavior, as this spurs consumers to refrain from temporal discounting of costs and benefits. The relative strength of a SNS usage habit

compared to a calculus’ vicinity of costs and benefits would in this context be a relevant case of future research. Altogether, it is suggested to further study personalization boundaries and its relation to continuous usage; a potential mediation role of the privacy calculus should not be shelved.

(25)

25

References:

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, 34-49. doi: 10.1016/j.jretai.2014.09.005 0022-4

Aguirre, E., Roggeveen, A. L., Grewal, D. & Wetzels, M. (2016). The personalization-privacy paradox: implications for new media. Journal of Consumer Marketing, 33(2), 98-110. doi: 10.1108/JCM-06-2015-1458

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, 305-321. doi: 10.1007/s11002-008-9056-z

Barnes, S. J. & Böhringer, M. (2011). Modeling use continuance behavior in microblogging services: the case of twitter. Journal of Computer Information Systems, 51, 1-10. Basak, E. & Calisir, F. (2015). An empirical study on factors affecting continuance intention of using Facebook. Computers in human behavior, 48, 181-189.

doi: 10.1016/j.chb.2015.01.055

Bleier, A. & Eisenbess, M. (2015). The Importance of Trust for Personalized Online Advertising. Journal of Retailing, 91(3), 390-409.

doi: 10.1016/j.jretai.2015.04.001

CBS. (2017, januari 23). CBS: Een op de zes jongeren zegt verslaafd te zijn aan sociale media. [press release]. Retrieved from http://www.cbs.nl

(26)

26 Chellappa, R. K. & Sin, R. G. (2005). Personalization versus privacy: An empirical examination of the online consumer’s dilemma. Information technology and

management, 6, 181-202. doi: 10.1007/s10799-005-5879-y

Chen, R. (2013). Members of social networking sites – an empirical examination. Decision

support systems, 54(3), 1219-1227. doi: 10.1016/j.dss.2012.10.028

Davies, S. (1997). Re-engineering the right to privacy: how privacy has been transformed from a right to a commodity. Technology and Privacy, 143-165.

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, 124-134. doi: 10.1080/15252019.2015.1082450

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, 368-383. doi: 10.1111/jcc4.12163

Dijkstra, A. (2014). The Psychology of Tailoring-Ingredients in Computer-Tailored Persuasion. Social and Personality Psychology Compass, 2, 765-784. doi:

10.1111/j.1751-9004.2008.00081.x

Ernst, P. (2014). Risk Hurts Fun: The influence of perceived privacy risk on social network site usage. In P. Ernst, Factors Driving Social Network Site Usage (pp. 45-56). Frankfurt

am Main, Germany: Springer Gabler. doi: 10.1007/978-3-658-09918-3_4

Hadija, Z., Barnes, S. B. & Hair, N. (2012). Why we ignore social networking advertising.

Qualitative Market Research: An International Journal, 15, 19-32. doi:

10.1108/13522751211191973

Hallam, C. & Zanella, G. (2016). Online self-disclosure: The privacy paradox explained as temporally discounted balance between concerns and rewards. Computers in Human

Behavior, 68, 217-227. doi: 10.1016/j.chb.2016.11.033

Hawkins, R. P., Kreuter, M., Resnicow, K., Fishbein, M. & Dijkstra, A. (2008). Understanding tailoring in communicating about health. Health education research,

(27)

27 Hayes, A. F. (2013). Introduction to mediation, moderation, and conditional process

analysis: A regression-based approach. New York, NY: Guilford Press

Ho, S. J. & Bodoff, D. (2014). The effects of web personalization on user attitude and behavior an integration of the elaboration likelihood model and consumer search theory.

Management Information Systems, 38(2), 497-520.

James, T. L., Warketin, M. & Collignon, S. E. (2015). A dual privacy decision model for online social networks. Information & Management, 52, 893-908.

doi: 10.1016/j.im.2015.07.010

Jung, E. J., McKnight, D. H., Jung, E. & Lankton, N. K. (2011). The Surprising Lack of Effect of Privacy Concerns on Intention to Use Online Social Networks. AMCIS Proceedings

17, 1-13. Retrieved from: http://aisel.aisnet.org/amcis2011_submissions

Kalyanamaran, S. & Sundar, S. S. (2006). The Psychological Appeal of Personalized Content in Web Portals: Does Customization Affect Attitudes and Behavior? Journal of

Communication, 56, 110-132. doi: 10.1111/j.1460-2466.2006.00006.x

Kehr, F., Wentzel, D. & Mayer, P. (2013). Rethinking the Privacy Calculus: On the Role of

Dispositional Factors and Affect. Paper presented at the thirty-fourth International

Conference on Information Systems, Milan, Italy, pp. 1-10.

Kelly, L., Gayle, K. & Drennan, J. (2010). Avoidance of Advertising in Social Networking Sites: The Teenage Perspective. Journal of Interactive Advertising, 10(2), 16–27. doi: 10.1080/15252019.2010.10722167

Krasnova, H. & Veltri, N. F. (2010). Privacy Calculus on Social Networking Sites: Explorative Evidence from Germany and USA. IEEE Conference publications 43, 1-10.

doi: 10.1109/HICSS.2010.307

Krasnova, H., Kolesnikova, E. & Günther, O. (2010b). Leveraging Trust and Privacy

Concerns in Online Social Networks: an Empirical Study. ECIS 2010 proceedings, 18, 1-12. Retrieved from: http://aisel.aisnet.org/ecis2010/

Ku, Y., Chen, R. & Zhang, H. (2013). Why do users continue using social networking sites? An exploratory study of members in the United States and Taiwan. Information &

(28)

28 Leshikar, E. D., Dulas, M. R. & Duarte, A. (2015). Self-referencing enhances recollection in both young and older adults. Aging, Neuropsychology and Cognition, 22, 388-412. doi: 10.1080/13825585.2014.957150

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

Li, Y. (2012). Theories in online information privacy search: A critical review and integrated framework. Decision Support Systems, 54. 471-481. doi: 10.1016/j.dss.2012.06.010

Li, Y., Liu, W., Jin, F. (2014). Research of the Relationship between Privacy Concern and Continuous Use Behavior in Online Personalized System. PACIS 2014 proceedings

18, 1-13. Retrieved from http://aisel.aisnet.org/pacis2014

Louise, K., Kerr, G. & Drennan, J. (2010). Avoidance of Advertising in Social Networking Sites: The Teenage Perspective. Journal of Interactive Advertising, 10 (2), 16-27 Maslowska, E. H., Smit, E. G. & Van Den Putten, B. (2011). Examining the (in) Effectiveness

of Personalized Communication. European advertising academy, 10th International Conference on Research in Advertising.

Maslowska, E. H., Smit, E. G. & Van Den Putten, B. (2016). It Is All in the Name: A Study of Consumers’ Responses to Personalized Communication. Journal of Interactive

Advertising, 16, 74-85. doi: 10.1080/15252019.2016.1161568

McKnight, D. H., Lankton, N. & Tripp, J. (2010). Social networking information disclosure and continuance intention: A disconnect. IEEE Conference publications 44, 1-31. doi: 10.1109/HICSS.2011.379

Nationaal Social Media Onderzoek (2017). [Statistics on Dutch social media use]. Newcom Research & Consultancy. Retrieved from http://www.newcom.nl/socialmedia2017 Oenema, A., Tan, F. & Brug, J. (2005). Short-term Efficacy of a Web-based Computer-Tailored Nutrition Intervention: Main Effects and Mediators. Annals of Behavioral

(29)

29 O’Donnell, K. & Cramer, H. (2015). People’s Perceptions of Personalized Ads. WWW 2015

Companion 24, 1293-1298. doi: 10.1145/2740908.2742003

Park, J. (2014). The effects of personalization on user continuance in social network sites.

Information Processing and Management, 50, 462-475.

doi: 10.1016/j.ipm.2014.02.002

Pavlou, P. A., Liang, H. & Xue, Y. (2006). Understanding and Mitigating Uncertainty in Online Exchange Relationships: A Principal-Agent Perspective. MIS Quarterly, 31, 105-136.

Sheeran, P. (2002). Intention-Behavior relations: A Conceptual and Empirical Review.

European Review of Social Psychology, 12, 1-36.

doi: 10.1080/14792772143000003

Smallwood, B. (2016). Resisting the siren call of popular digital media metrics: Facebook research shows no link between trendy online metrics and ad effectiveness. Journal of

Advertising Research, 56(2), 126-131

Smith, H. J., Dinev, T. & Xu, H. (2011). Information Privacy Research: An Interdisciplinary Review. MIS Quarterly, 35(4), 989-1015.

Sun, Y., Liu, L., Peng, X., Dong, Y. & Barnes, S. J. (2013). Understanding Chinese user’s continuance intention toward online social networks: an integrative theoretical model.

Electron Markets, 24, 57-66. doi: 10.1007/s12525-013-0131-9

Tucker, C. E. (2014). Social Networks, Personalized Advertising, and Privacy Controls.

Journal of marketing research, 51, 546-562. doi: 10.1509/jmr.10.0355

Van Doorn, J. & Hoekstra, J. C. (2013). Customization of online advertising: The role of intrusiveness. Marketing letters, 24(4), 339-351. doi: 10.1007/s11002-012-9222-1 Walrave, M., Poels, K., Antheunis, M. J., Van den Broek, E. & Van Noort, G. (2016). Like or

dislike? Adolescents’ responses to personalized social network site advertising.

Journal of marketing communications, 1-18. doi: 10.1080/13527266.2016.1182938

White, T. B., Zahay, D. L., Thorbjornsen, H. & Shavitt, S. (2008). Getting too personal: Reactance to highly personalized email solicitations. Marketing letters, 19, 39-50. doi: 10.1007/s11002-007-9027-9

(30)

30 Wood, W. & Neal, D. T. (2007). A New Look at Habits and the Habit-Goal Interface.

Psychological Review, 114, 843-863. doi: 10.1037/0033-295X.114.4.843

Yin, G., Cheng, X. & Zhu, L. (2011). Understanding Continuance Usage of Social Networking Services: A Theoretical Model and Empirical Study of the Chinese Context.

Online Communities and Digital Collaborations 32, 1-13. Retrieved from

http://aisel.aisnet.org/icis2011

Zhou, T. & Li, H. (2014). Understanding mobile SNS continuance usage in China from the perspectives of social influence and privacy concern. Computers in Human Behavior,

37, 283-289. doi: 10.1016/j.chb.2014.05.008

Zhu, H., Ou, C. X. J., Van den Heuvel, W. J. A. M. & Liu, H. (2016). Privacy calculus and its utility for personalization services in e-commerce: An analysis of consumer decision-making. Information & Management. doi: 10.1016/j.im.2016.10.001

Appendices

Table A1

Means for the 14 pretest items of Facebook information (N = 18)

Pleasant – unpleasant (1-7) Useful – useless (1-7) Acceptable – unacceptable (1-7) Impersonal - personal (1-7) M M M M Maverage First name 4.61 4.44 3.56 5.61 4.56 Surname 5.61 4.94 4.67 5.89 5.28 Gender 3.56 2.83 2.67 4.00 3.27 (low) Address 6.39 5.67 6.11 6.06 6.06 Email Address 6.17 5.50 5.89 5.78 5.84

Cell phone number 6.78 6.06 6.61 6.22 6.42 (high)

Profile picture 5.39 4.61 5.11 5.83 5.24

Album picture 5.89 5.22 5.83 5.50 5.61

Date of birth 5.33 3.44 4.56 5.50 4.71 (moderate)

Friends 4.78 3.61 4.11 5.22 4.43

Interests 3.50 2.33 2.61 5.06 3.38

Status updates 5.39 4.44 5.33 6.00 5.29

(31)

31

School information 4.89 4.11 4.22 4.89 4.53

Table A2

Means for the three manipulation check items on Facebook information (N = 194) and the experimental conditions (N = 87) Pleasant – unpleasant (1-7) Useful – useless (1-7) Acceptable – unacceptable (1-7) Impersonal – personal (1-7) M M M M Maverage Mexpconditions Gender 3.94 3.71 3.30 4.05 3.75 (low) 3.68

Cell phone number 6.51 6.22 6.06 6.10 6.22 (high) 5.91

Date of birth 4.91 4.25 4.48 5.20 4.71 (moderate) 4.35

Appendix B1: Facebook screenshot of the low personalization condition (Gender, man).

(32)

32 Appendix B3: Facebook screenshot of the moderate personalization condition (Date of birth).

Appendix B4: Facebook screenshot of the high personalization condition (Cell phone number).

(33)

33 Appendix C1: Questionnaire items for mediator Perceived benefits.

1. The use of Facebook helps me cope with negative situations I experience. 2. The use of Facebook helps me express my personality and feelings.

3. The use of Facebook helps me learn new things or think about things in a new way. 4. The use of Facebook does not help me control how others see me.

5. The use of Facebook does not help me make new social or business contacts.

6. The use of Facebook does not help me contribute to things that I think are important. 7. The use of Facebook does not help me influence people in my social circle.

8. The use of Facebook helps me better understand my stances on problems or situations.

9. The use of Facebook helps me come in contact with others that share my views and interests, and maintain this contact.

10. The use of Facebook helps me feel more connected with others.

(34)

34 1. Compared with other topics, privacy is important to me.

2. I feel very concerned about my privacy.

3. I do not worry about my privacy as a result of using Facebook.

4. The danger to people’s privacy when they use social networking sites like Facebook has been overblown.

5. I am concerned that information I share with Facebook could be misused.

6. I am concerned about sharing information with Facebook, because it could be used in a way I did not foresee.

Appendix D: Elaboration on the manipulation check filter process.

The first manipulation check filter question was included to measure recall of the advertisement: “Earlier in this study you saw an image of a Facebook profile. Did you see an advertisement on it?”. When answered with yes, respondents received the question “In how far did you perceive the advertisement on the Facebook page as personal?” which could be answered on a seven-point Likert scale ranging from ‘very personal information (1)’ to ‘very

impersonal information (7)’. The last question, included to measure recognition of the type of

advertisement, was “What type of information was the advertisement on the Facebook page based?”. Respondents could answer this question with ‘Gender’, ‘Cell phone number’, ‘Date

of birth’ or ‘Did not see, don’t know’. Results on the first filter question revealed that 11

respondents (5.7%) did not remember seeing an advertisement at the time of the question. Manipulation check item 2 showed that respondents in both Gender and Date of Birth conditions perceived the information in the corresponding advertisements as rather impersonal information, whereas the Cell phone number condition was perceived as more

Threat or treat : personalized social network site advertising and consumer continuance intention