cgchgc
Utz, S., & Kramer, N. (2009). The privacy paradox on social network sites revisited: The role of individual characteristics and group norms. Cyberpsychology: Journal of Psychosocial Research on Cyberspace, 3(2), article 2.
http://cyberpsychology.eu/view.php?cisloclanku=2009111001&article=2

The privacy paradox on social network sites revisited: The role of individual characteristics and group norms

Sonja Utz1, Nicole C. Kramer2
1 VU University Amsterdam, the Netherlands
2 University of Duisburg-Essen, Germany

Abstract

Users of social network sites (SNS) often state that they are concerned about their privacy, yet they often disclose detailed personal information on their profiles. This paper assessed the privacy settings of users of two large European SNS. More importantly, it also examined which factors predict the choice of specific privacy settings. The main focus was on the trade-off between privacy concerns and impression management. The paper also looked at the role of the dispositional variables trust and narcissism. These individual factors were contrasted with the effects of perceived group norms. Across three studies it was found that the vast majority of users protected at least certain parts of their profile (e.g., pictures, email address). Moreover, higher protection of profiles was consistently predicted by greater privacy concerns. Impression management motives and narcissism led to less restrictive privacy settings, but these results were less consistent across studies. Perceived social norms played a role in both SNS, whereas dispositional trust had no effect.

Keywords: social network sites, privacy settings, privacy concerns, social norms

Introduction

Social network sites (SNS), profile sites that also display the connections between users, have become increasingly popular over the last years (boyd, 2006, boyd & Ellison, 2007). Unlike former anonymous modes of computer-mediated-communication such as chats or online role-playing games, these sites have been described as “nonymous” environments in which people are related to each other by “anchored relationships” (Zhao, Grasmuck, & Martin, 2008). This entails that the users can be identified as well as found in real life. Given this, it is important to analyze how people deal with their personal data. Indeed, it has been shown that users of SNS state that they are worried about their privacy, but put at the same time detailed personal information on their profiles. This phenomenon has been described as the privacy paradox (Barnes, 2006). The dangers of putting too much personal information online have received a great deal of attention in the media in recent times (cf. Lewis, Kaufman, & Christakis, 2008); and SNS have reacted by offering more fine-grained privacy settings (Berger, 2006). Thus, it is time to revisit the privacy paradox on SNS. Lewis et al. (2008) predicted a movement towards more restrictive privacy settings. Users have to find a balance between privacy concerns and impression management goals. The aim of the present paper is to examine how SNS users deal with privacy and which factors predict the choice of specific privacy levels.

The privacy paradox
Research on SNS and other social media found a paradox, a discrepancy between privacy concerns and actual privacy settings (Barnes, 2006). Analyses of profiles have found that SNS users provide a large amount of personal information on public profiles. Gross and Acquisti (2005) analyzed the Facebook profiles of more than 4000 students and found that only a small percentage had changed the default privacy settings. Thelwall (2008) analyzed more than 20,000 MySpace profiles and found that only 27% were set to private. In 2007, Lewis et al. (2008) downloaded the Facebook profiles of a whole class of a private American university and found that only one third was set to private. A similar percentage has been reported in a study on Hyves users, also conducted in 2007 (Utz, 2008).

Although alarming, these numbers should be seen as snapshots. SNS are rapidly growing and evolving, and the privacy issue has received much attention, both from the media and on the sites themselves (see, for example, the protest of Facebook users after the introduction of the mini-feeds, or the Beacon program; Story & Stone, 2007). The sites now offer more sophisticated settings, and users are learning. Many teens are aware of the risks of putting information online and make careful decisions about what information to put online under which circumstances (Lenhart & Madden, 2007). Similarly, Livingstone (2008) derived from in-depth interviews with teenagers that they, in general, are not unconcerned about their privacy. However, today's teenagers display easily personal information that previous generations often have regarded as private (e.g. age, politics, income, religion, sexual preference). The teenager's definition of privacy is not tied to the disclosure of certain types of information, but– for them – privacy is assured as soon as they have control over who knows what about them (Livingstone, 2006). This is in line with Stein and Sinha's (2002) definition of privacy as ‘the rights of individuals to enjoy autonomy, to be left alone, and to determine whether and how information about one’s self is revealed to others’ (p. 414). Although these considerations might lead to the conclusion that the privacy paradox actually is no paradox but due to different perspectives on privacy, Livingstone (2008) also showed that the teenagers struggle with the privacy settings of the SNS. Partly due to limited internet literacy and partly due to poorly designed site settings, the users are unsure how to control who can see what about them.

Other research suggests that the usage of privacy settings is not first and foremost a matter of capabilities but of subjective decisions which are influenced by various variables such as media coverage or privacy concerns: Lewis et al. (2008) propose that the commercialization of SNS and the reports about privacy intrusions will lead to more restrictive privacy settings. Based on an analysis conducted in December 2008, Lenhart (2009) reports that 60% of adult SNS users had changed the default settings (visible for everyone) to more restrictive settings. Thus, there might indeed be a movement towards more restrictive privacy settings. These recent studies are all based on American Internet users. The present paper will contribute to the generalizability of these findings by examining the privacy settings in two European SNS.

Hyves, the most popular Dutch SNS, and StudiVZ, the SNS most popular among German students, were selected as research sites. Both are leisure-oriented SNS. StudiVZ was founded in 2005 and is like Facebook in it's early days restricted to students. Later, a version open to everyone (MeinVZ) was founded. Students who have finished their university education can migrate with their profiles to MeinVZ, and profiles can be available for both communities. The design of StudiVZ is similar to the one of Facebook; it is even so similar that Facebook (unsuccessfully) sued StudiVZ (Kleinz, 2009).

Hyves has been founded in 2004 and it has meanwhile more than 9 million registered users. The basic functions are like on other SNS: people create profiles and maintain relationships. The friends are displayed more prominently than on Facebook. There are not so many applications and social games as on Facebook, but people can share pictures and videos, and every user has a space to write a blog. The default privacy settings are that every member of Hyves can see the profiles, but users can adjust this per part of the profile (e.g. pictures, comments from other users, email address, relationships status) to "nobody", "only friends", "friends of friends", "Hyvers" or "everybody". The privacy options on StudiVZ are similar: "only friends", "friends of friends", "all people at my university, friends, and friends of friends", and "everybody". However, the settings are less fine-grained. Choices are not made for each part of the profile, but for the profile as a whole.

These two sites were chosen because they are both very popular in the respective countries. However, they differ also in the user demographics. StudiVZ is restricted to students, whereas Hyves is open for everybody. To see whether there is indeed a broad privacy movement and not just a change of norms in small niche SNS, it is important to compare different SNS. The first research goal of the paper is to gain more knowledge about the privacy settings in these SNS; the first research question is therefore:

RQ1: What are the actual privacy settings of European SNS users?

To determine whether there is still a paradox, it is also important to measure the privacy concerns of users, and to examine whether the privacy concerns predict the privacy settings. The paper will not only examine the predictive role of privacy concerns, but also the role of other variables that might influence the choice of privacy settings: impression management, dispositional variables, and perceived social norms (Study 2 and 3).

The privacy paradox might arise because users have to find a trade-off between two opposing motives: privacy concerns and impression management. The two main purposes of a SNS are self presentation and maintaining relationships. A user can only be found by old acquaintances or friends if the profile is visible. Impressing potential employers or dates also requires a public profile. Indeed, impression management has been found to be a major motive for hosting a profile on a SNS (Kramer & Winter, 2008; Banczyk, Kramer, & Senokozlieva, 2007). Zhao et al. (2008) showed by means of a content analysis of Facebook accounts that in the nonymous environment of the SNS the users produced identities especially by implicit identity claims (such as photos, groups or quotes) instead of explicit descriptions of their person. Also, the Facebook descriptions resembled highly socially desirable identities individuals aspired to have offline but have not yet been able to embody in real life (in contrast to “true selves” presented in anonymous environments and “real selves” shown in face-to-face interaction). Similarly, Siibak (2009) found that SNS users were very strategic and deliberately selected pictures with the goal to appear more popular on the SNS.

Thus, impression management is an important concern for SNS users. Tufekci (2009) argued that privacy has to be seen as a compromise between pressures for withdrawal and disclosure. Therefore, opposite effects are expected for privacy concerns and impression management.

H1: The higher the privacy concerns, the stricter the privacy settings.

H2: The more the Internet is used for impression management, the less strict are the privacy settings.

Trust and narcissism are dispositions that might influence privacy settings because they correspond to the opposing motives that might influence the choice of privacy settings: privacy protection vs. impression management in front of an audience. Dispositional trust is an individual's inclination to believe that others will act in his or her best interest (Kramer, 1999). Individuals with a high propensity to trust other people also have more trust in various e-commerce venues (Gefen, 2000; McKnight, Choudhury, & Kacmar, 2002). For SNS, positive correlations were found between trust in the SNS and revealing home town and cell phone number, as well as between trust in other members and revealing one’s email address and relationship status (Dwyer et al., 2007). Christofides, Muise, and Desmarais (2009) reported negative effects of trust on information control on Facebook. This leads to the following hypothesis:

H3: The higher the dispositional trust, the less strict the privacy settings.

It has often been proposed in the media and popular science that narcissism, the feeling of being a very special person who deserves a lot of attention, might be the reason for self-exhibition on social media such as SNS or YouTube (Keen, 2007; Orlet, 2007; Rosen, 2007). A recent study found relationships between narcissism and social activity on Facebook and self-promoting content (Buffardi & Campbell, 2008). According to Ibrahim (2009, p. 82), SNS signify a narcissistic turn in self-presentation. Sharing and exchanging personal information between friends and even strangers is an important element of SNS. Ibrahim argues that the benefits of revealing personal data to strangers are often perceived as higher than the risks associated with this behavior. Therefore, it can be expected that people high on narcissism favor a large audience and make their profile visible for many users.

H4: The higher narcissism, the less strict the privacy settings.

Study 1

Method

Participants
An online survey was conducted among the members of Hyves. The questionnaire was put online in June 2008 and announced in several Hyves groups and via banners. Participants could win 15- € in gift cards. The announcement did not mention that the study was about privacy, so it is unlikely that mainly Hyvers interested in privacy issues participated. 271 individuals started to fill in the questionnaire, and 217 (62 males, 155 females) completed it to the end. The mean age of the respondents was 26 (SD = 9.86), which corresponded to the mean age of the average user at that time (according to the official Hyves statistics which are available for gold members). The majority (58%) attended school or university, while 34% were employed. On average, participants had been members of Hyves for 20 months (SD = 12), and most of them logged in several times a day (51%) or (almost) daily (27%).

Measures
Demographics. Age, gender, occupation, and education level were assessed.

Hyves activity. Duration of membership, frequency of visits, and number of friends were assessed.

Dispositional trust. Dispositional trust was measured with six items (Van Lange, Van Vugt, Meertens, & Ruiter, 1998). An example item is "Most people are trustworthy" (α = .76, M = 3.17, SD = .63).

Narcissism. Narcissism was measured by a shortened and modified version of the Narcissistic Personality Inventory (Raskin & Terry, 1988). Instead of the forced-choice format, respondents indicated on five-point scales from not at all to very much to what extent five statements such as "I am a special person" applied to them (α = .75, M = 2.86, SD = .70).

Privacy concerns. The scale privacy concerns consisted of five items such as "Sometimes I find it scary how much information about me is available on the Internet" (α = .77, M = 3.31, SD = .90).

Impression management. Using the internet for impression management consisted also of five items such as "I use the Internet to influence my image" or "I like the Internet because more people can notice me/my profiles" (α = .76, M = 2.45, SD = .77).

Privacy settings. First, respondents were asked whether they had ever changed their default privacy settings. Individuals who had changed their privacy settings indicated which audience could see various parts of their profile. The answer categories were "nobody", "friends", "friends of friends", and the combined options "Hyvers/everybody". "Nobody" means that the user had not even filled in this information. Eight options were provided: last name, email address, cell phone number, relationship status, krabbels (comments on the profile), blogs, photos, and wiewatwaar (whowhatwhere – a feature that allows the user to enter what he or she is doing, where, at that very moment, e.g. writingapaper@office). Answers were coded such that higher values represented more protective privacy settings, and were averaged across the eight options (α = .89, M = 2.79, SD = .74).

Results

Descriptive analysis
Table 1: Privacy settings, Study 1
fig

Most respondents (74%) had changed their privacy settings. In Table 1, the actual privacy settings of the whole sample are displayed. The default settings are that everything is visible to all other Hyvers. Most Hyvers who had changed their privacy settings had changed them into more restrictive privacy settings. In particular, identifying information, i.e. cell phone number and email address, was visible only for friends or not even displayed at all. Photos, comments, and blogs were also often only visible for friends, but the option "friends of friends" was chosen more often here.

Predictors of privacy settings
To examine which of the dispositional and Internet-specific variables predict the choice for privacy settings a regression analysis was conducted. Gender, age, duration of Hyves membership and frequency of visits were included as control variables. The overall model was significant, F(8, 202) = 4.38, p < .001, R2 = .15. Neither the control variables nor the dispositional variables narcissism and trust had a significant effect. Only the Internet specific variables predicted actual privacy settings. Stronger privacy concerns were related to more restrictive privacy settings, β = .29, t(202) = 4.38, p < .001, and impression management was related to less restrictive privacy settings, β = -.22, t(202) = -2.97, p < .01. Hypotheses 1 and 2 were supported.

Discussion
Study 1 examined the privacy settings of Hyves users and found that most users have changed their privacy settings. Just seven months earlier, only 30% of the Hyves users had restricted the access to their profile (Utz, 2008). The two significant predictors of privacy settings were privacy concerns and impression management motives. Thus, the data show that privacy settings are indeed the results of an optimization of these two opposing needs.

Dispositional trust and narcissism did not influence the choice of privacy settings. This pattern can be explained by the correspondence principle – specific attitudes are better predictors of specific behavior (Ajzen & Fishbein, 1977). Study 1 focused on individual dispositions and attitudes, but changes in privacy settings might also be driven by group processes such as norms. If users encounter more and more often messages such as "unfortunately, this profile is not visible for everybody" whilst browsing the profiles of friends of friends, they might infer a norm regarding the privacy settings in the SNS. According to Social Impact Theory (Latane, 1981), an individual’s attitudes and behavior are influenced by the attitudes and behavior of others, especially close and important groups such as peers. The role of perceived social norms has been often studied and confirmed in the domain of undesirable behavior such as heavy drinking behavior (e.g., Berkowitz & Perkens, 1986; Pedersen, LaBrie, & Lac, 2008). However, perceived norms should also influence the privacy settings on SNS. Lewis et al. (2008) found that the number of friends with a private profile predicted the likelihood of a user having a private profile him- or herself. Profiles of friends are always visible, so the perceived norm derived from browsing profiles of users not in the network can be expected to have even more influence. Whereas Lewis et al. (2008) established a link between absolute number of friends with private profiles and privacy settings of the users, the present research will examine the link between perceived social norm and privacy settings.

H5: The more restrictive the perceived norms, the more restrictive the privacy settings.

A limitation of the first study is the self-selected sample of active Hyvers. Almost 80% logged in daily. Active Hyvers might be more concerned about their privacy because they are so active and place more sensitive information on their profiles. To avoid this possible bias, the second study used a convenience sample of students, who had come to the lab for an unrelated psychological experiment. The dispositional variables were not measured in this study because the experiment took more than half an hour; instead, the focus was on perceived social norms.

Study 2

Method

Participants
Seventy (27 males, 43 females) students, who came to the lab in September 2008 to participate in a social psychological experiment and had a Hyves account, participated in the study. The mean age of the participants was 21 (SD = 4.04). As in study 1, the mean duration of Hyves membership was M = 20 months (SD =10). However, only a minority of seven percent logged in several times per day. Fourteen percent logged in (almost) daily, whereas the majority of 41% visited Hyves several times per week. Twenty-three percent logged in once a week, and 14% less than once a week. Thus, a sample of less active Hyves users was actually reached.

Measures
Hyves activity. As in study 1, duration of membership and frequency of visits were assessed as control variables.

Privacy concerns. Privacy concerns (α = .72) were measured using the same items as in study 1. However, this time, answers were given on a 7-point scale (M = 4.53, SD = 1.07).

Impression management. Using the Internet for impression management was also measured with the same scale as in study 1. Answers were also given on a 7-point scale (α = .75, M = 3.33, SD = .97).

Privacy settings. Again, participants were asked whether they had actually changed their privacy settings. Additionally, they indicated how often (never, once, a few times, more often) and how long ago (last week, a few weeks ago, a few months ago, a long time ago, when I became a member) they had changed the privacy settings. To measure the actual privacy settings, the same categories as in study 1 were presented. Five answer categories were presented: "nobody", "only friends", "friends of friends", "Hyvers", "everybody”, and an additional "don't know/not applicable" category (for people who did not use, for example, the blog feature).

Perceived norm. To assess the perceived norm, participants were asked to estimate the percentage of their friends who made krabbels available only for friends. Next, they estimated the percentage of friends who made photographs available only for friends. The same estimates then had to be made for the average Hyver. An exploratory factor analysis showed that these variables loaded on one factors and the items were averaged (α = .85, M = 46, SD = 20).

Results

Descriptive results
Again, the majority (83%) had changed their privacy settings. Most of those (55%) had done so once, 43% several times and one person had done so more often. Thirty-four percent of the respondents had changed their privacy settings (on the last occasion) when they became a member of Hyves, 21% had done so a long time ago, 29% a few months ago, 14% a few weeks ago and 1% in the previous week. Thus, roughly 45% had changed their privacy settings within the last weeks or months, indicating that there is indeed a change in sensitivity to privacy issues over time.

Table 2: Privacy settings, Study 2
fig
As can be seen in Table 2, the results were quite similar to study 1. If there were differences, the sample in study 2 had more restrictive privacy settings. That is, even more information was not made available at all or only for friends. Again, email address and cell phone number were the most protected pieces of information – almost nobody gave this information to people other than friends.

Prediction of privacy settings
To examine whether perceived norms explain additional variance, a hierarchical multiple regression was conducted. Privacy concerns, using the Internet for impression management and the control variables gender, age, duration of Hyves membership and frequency of login were entered in the first block; the perceived norm was entered in the second block.

The first analysis was marginally significant, F(6,62) = 1.99, p < .10, R2= .16. Only privacy concerns predicted privacy settings, β = .29, t(66) = 2.09, p < .01. The second step explained an additional 10% of variance, F(2,60) = 4.21, p < .05 for F change. Privacy concerns were still a significant predictor, β = .35, t(60) = 2.81, p < .01. Hypothesis 1 is therefore again supported. In line with hypothesis 5, perceived norms regarding what to show only to friends were also a significant predictor, β = .33, t(60) = 2.66, p < .01. No other effects were significant, ts < 1.48.

Discussion

The sample of study 2 consisted of less active Hyves users. Nevertheless, a very similar pattern of privacy settings was found. Again, the vast majority of Hyvers had changed their privacy settings. Although about one third had done so when they became a member, a significant amount had changed the settings within the last weeks or months. Thus, the sensitivity for privacy issues is in a constant state of flux; many users have changed their privacy settings several times since becoming a member. More importantly, privacy concerns predicted the actual privacy settings, again indicating that the privacy paradox is slowly becoming resolved. Interestingly, using the Internet for impression formation was not significant in this analysis. Obviously, the behavior of less active Hyves users is less driven by impression management concerns.

More important, perceived social norm turned out to be an important predictor. Perceived norms were not significantly correlated to privacy concerns; one’s own attitude and social influence are therefore two independent processes influencing the choice of privacy settings.

The goal of study 3 was to replicate the findings in a different SNS in order to examine whether heightened privacy concerns and active privacy management is a more general phenomenon. StudiVZ was used as research site.

Study 3

Method

Participants
A paper and pencil questionnaire was handed out to the students of a social psychology course in December 2008. One hundred and forty seven (54 males, 93 females) StudiVZ users participated in this study. The mean age of the respondents was 21 (SD = 1.99). On average, they had been a member of StudiVZ for 19 months (SD = 8.7).

Measures
Dispositional trust and narcissism. Dispositional trust and narcissism were measured with the same scales as in study 1 (dispositional trust: α = .72, M = 2.65, SD = .60; narcissism: α = .75, M = 3.09, SD = .66).

Privacy concerns. The scale privacy concerns consisted of three items also used in the first two studies (α = .73, M = 2.49, SD = 1.18).

Privacy settings. StudiVZ has different options for the privacy settings, meaning that some of the questions had to be adjusted. StudiVZ is closely linked with MeinVZ, and profiles can be visible in both SNS. Thus, users of StudiVZ can decide about the visibility of their profile within StudiVZ and within MeinVZ. The options within StudiVz are “only friends”, “friends and friends of friends”, “all people at my university, friends and friends of friends” and “everybody”. The options within MeinVz are basically the same; only the option “all people at my university” is missing. These two items were scored such that higher values indicated more privacy and were combined into one privacy index (α = .93, M = 2.78, SD = 1.38). As in study 2, respondents were also asked how often they had changed their privacy settings.

Perceived norm. To assess the perceived norm, participants were asked to estimate the percentage of their friends and the percentage of average StudiVZ users who made their profile available only for friends. These two items formed the variable norm (α = .73, M = 49, SD = 19).

Results

Descriptive analyses
Eighty-nine percent of the StudiVZ users had changed the default privacy settings. Twenty-nine percent had done so only once, the vast majority of the respondents, 51%, had done so several times, and nine percent had changed their privacy settings even more often. Seventeen percent had changed their privacy settings when they became a member. Nine percent had changed their privacy settings within the last week, 30% within the last few weeks and 33% within the last months.

Table 3: Privacy settings, Study 3
fig
As can be seen in Table 3, about half of the StudiVZ users made their profile only available for friends. About 11% made their profile also available for friends of friends.

Prediction of privacy settings
A hierarchical regression analysis was conducted: Gender, duration of membership, dispositional trust, and narcissism were entered in block 1, privacy concerns in block 2 and perceived norm in block 3.

The first step revealed a significant model, F(4,138) = 3. 57, p < .05, R2 = .09. Narcissism was the only significant predictor, β = -.20, t(138) = -2.31, p < .05. The higher the narcissism score, the less strict the privacy settings. Entering privacy concerns revealed a significant model, F(5,137) = 4.65, p < .001, and resulted in a significant increase in explained variance to R2 = .15, F change (1,137) = 8.23, p < .001. Privacy concerns had a significant effect, β = .23, t(137) = 2.87, p < .01, and narcissism remained a significant predictor, β = -.19, t(137) = -2.18, p < .05. Norms, entered in step 3, explained an additional nine percent of variance, F change (1,136) = 15.96, p < .001. More restrictive norms predicted more restrictive privacy settings, β = .31, t(136) = 4.00, p < .01. Privacy concerns, β = .21, t(136) = 3.01, p < .01, and narcissism, β = -.16, t(136) = -1.97, p = .05 remained significant. Hypotheses 1, 4 and 5 were supported.

Discussion
Again, roughly 90% of the respondents had changed their privacy settings. Thus, the basic finding that SNS users care about their privacy and adjust their privacy settings can be generalized to StudiVZ, another large SNS. Again, many users had changed their privacy settings just recently, often only during the last weeks. As in the first two studies, privacy concerns predicted the actual privacy settings of the profile, indicating that the gap between privacy concerns and actual privacy-protecting behavior is closing. Moreover, the study replicated the finding that norms play an important role.

Narcissism predicted the protection of the profile. Impression management motives were not measured in this study, but narcissism is also related to presenting oneself favorably in front of a large audience. Thus, this result shows again that users have to find a trade-off between opposing motivations.

General Discussion

This paper revisited the privacy paradox by studying the privacy settings of users of two popular European SNS, Hyves and StudiVZ. The first sample was a self-selected sample of primarily very active Hyvers; the second and third were student samples also including less active SNS users. Across all three studies, it was found that the vast majority of users had changed the default privacy settings into more restrictive settings. One year earlier, only 30% of Hyves users had done so (Utz, 2008). This could be due to this earlier sample, but studies 2 and 3 found that many users had changed their privacy settings only recently. Thus, sensitivity to privacy issues seems to have increased. Earlier studies reported a privacy paradox, with users stating that they were concerned about their privacy, but not protecting their profiles (Bosau, Fischer, & Koll, 2008; Dwyer et al., 2007; Tufekci, 2008). In the present studies, most users protected their profiles, and, as can be seen in Table 4, privacy concerns consistently predicted the choice of privacy settings.

Table 4: Summary of the results across studies
fig
The data support the notion of a privacy movement suggested by Lewis et al. (2009). When SNS were still a niche phenomenon, users enthusiastically put a lot of sensitive information on their profiles. When SNS became mainstream, the first negative consequences occurred. People were fired or did not get a job because of party pictures on Facebook. The word about these events spread, within the SNS, but also in the media. As a consequence, awareness of the dangers grew. At the same time, users got more experienced and learned how to deal with new technologies.

It has been argued before that privacy settings arise from an optimization between competing pressures for disclosure and withdrawal (Tufekci, 2008). Impression management is an important motive in the nonymous environment of SNS (Kramer & Winter, 2008; Zhao et al., 2008). The present research showed that there is indeed a trade-off between opposing motivations. Whereas stronger privacy concerns resulted in more restrictive profiles, impression management motives (study 1) and narcissism (study 3) resulted in less restrictive privacy settings. The present research extended this trade-off model by including the roles of norms. Study 2 and 3 showed that norms explained an additional portion of variance. Perceived norms influence many aspects of student behavior, such as drinking and smoking behavior. The present results show that perceived norms play also an important role when it comes to the appropriate use of SNS. At least within Hyves and StudiVZ, a norm of making profiles only visible for friends or friends of friends has emerged.

Interestingly, the personality traits did not play an important role. Dispositional trust had no significant effect, and narcissism only predicted the privacy settings of the StudiVZ sample. The first is in contrast with the findings by Christofides (2009). This discrepancy can be resolved by considering the correspondence principle (Ajzen & Fishbein, 1977). Christofides et al. (2009) measured trust, but not more specific privacy concerns. General attitudes are known to be worse predictors; indeed, their model explained only 4% of the variance. Privacy concerns, a more specific measure, were a reliable predictor that explained more variance. Future research should focus on more specific models to explain the choice of privacy settings.

A reason for the inconsistent findings on narcissism might be the different operationalizations of the privacy measure due to the different options provided by the two SNS. In StudiVZ, privacy settings are set for the whole profile, whereas Hyves offers more fine-grained settings. Narcissism did not predict protection of cell phone number or email address – the aspects included in the Hyves privacy measure. However, it predicted making the whole profile public, that is, self-presentation in front of a large audience. Ibrahim (2009) related the narcissistic turn of SNS to the exchange of self-promoting information, and several authors showed that photos are used to construe a popular impression (Siibak, 2009; Zhao et al.. 2008). Impression management had an effect in the sample of very active users, indicating that active users use SNS for impression management and not only for the maintenance of relationships. When less active users are included in the sample, privacy concerns and norms became more important.

We would like to note some practical implications of the studies. First, parents and teachers do not have to be too worried about the behavior of students on SNS. Privacy concerns were a consistent predictor of actual privacy settings. Educational messages should therefore further point to potential dangers of public profiles. Norms also played an important role. Adolescents and young adults are especially sensitive for peer pressure. Providing information about the privacy-protecting behavior of peers could also lead to more restricted profiles. Moreover, a campaign might be more successful if the message is delivered by a peer than by a teacher or spokesperson.

There are also some limitations of the current studies. Self-reports were used to assess the privacy settings. It would be better to have actual behavioral data. However, since privacy settings have become more fine-grained, it is not possible to observe them. The researcher can see whether a profile is restricted, but does not know whether it is available only for friends or also for friends of friends. It is also not possible to draw a random sample of users in Hyves because only people who have entered their last name can be found via the search machine. If a profile is not visible, the friends are also not visible. Sampling by clicking for example on the 10th friend would always stop when a protected profile is encountered and public profiles would be overrepresented. Thus, relying on self reports is the most adequate method.

It remains problematic to claim that the high numbers of people protecting their profile indeed represent a privacy movement, because all three studies were cross-sectional and the study conducted in 2007, in which only 30% protected their profile, used a different sample (Utz, 2008). A longitudinal study would be the more appropriate test. However, we found similar patterns in different communities and among active and less active users. Moreover, the fact that many respondents indicated that they had changed their privacy settings recently also strengthens the claim that choosing more restrictive privacy settings is a recent trend.

The present research revealed interesting results, but it certainly does not provide the last word in the privacy debate. As a consequence of the restriction of profiles, users are now frequently even unable to view the profiles of friends of friends. This leads to frustration, as one of the respondents noted in the open text field provided at the end of the survey. Thus, it could be that users start to open their profiles for friends of friends in the near future. At the same time, SNS are continuously working on more fine-grained privacy settings; it remains open for future research how these new options will be used. Choosing the optimal privacy settings continues to be a dynamic process. However, the present research has clearly shown that users of SNS are actively dealing with the privacy question.

References

Ajzen I. & Fishbein M. (1977). Attitude-behavior relations: A theoretical analysis and review of empirical research. Psychological Bulletin, 84, 888-918.

Banczyk, B., & Kramer, N. C. & Senokozlieva, M. (2008). “the wurst” meets “fatless” in MySpace. The relationship between self-esteem, personality and self-presentation in an online community. Paper presented at the Congress of the International Communication Association, May 2008, Montreal, Canada.

Barnes, S.B. (2006). A privacy paradox: Social networking in the United States. First Monday, 11, Retrieved August 8, 2008 from http://firstmonday.org/htbin/cgiwrap/bin/ojs...

Berger D. (2006, September 9). Facebook get the hint, adds privacy setting to News Feed. Retrieved August 8, 2008 from www.gadgetell.com/tech/comment/facebook-gets.../

Berkowitz, A.D., & Perkins, H.W. (1986). Problem drinking among college students: A review of recent research. Journal of American College Health, 35, 21-28.

Bosau, C., Fischer, O., & Koll, M. (2008, July). StudiVZ - Determinants of social networking and dissemination of information among students. Paper presented at the International Congress of Psychology, Berlin (Germany).

boyd, d. (2006) ‘Friends, friendsters, and Top 8:Writing community into being on social network sites’ First Monday, 11, Retrieved March 9, 2007, fromhttp://firstmonday.org/htbin/cgiwrap/bin/ojs...

boyd, d., & Ellison, N. (2007) ‘Social Network Sites: Definition, History, and Scholarship’, Journal of Computer-Mediated Communication 13(1), Retrieved January 30, 2008 from http://jcmc.indiana.edu/vol13/issue1/boyd.ellison.html

Buffardi L.E., & Campbell, W.K. (2008) Narcissism and social networking web sites. Personality and Social Psychology Bulletin, 34, 1303-1314.

Christofides, E., Muise, M., & Desmarais, S. (2009). Information disclosure and control on Facebook: Are they two sides of the same coin or two different processes? CyberPsychology & Behavior, 12, 341-345.

Dwyer C., Hiltz, S. R., & Passerini K. (2007). Trust and privacy concern within social networking sites: A comparison of Facebook and MySpace. AMCIS Proceedings 2007: Paper 339.

Garfinkel, S. (2000). Database nation: The death of privacy in the 21st century. Sebastopol, Calif.: O'Reilly.

Gefen, D. (2000). E-commerce: the role of familiarity and trust. Omega: The International Journal of Management Science, 28, 725–37.

Gross, R., & Acquisti A. (2005). Information revelation and privacy in online social networks. Alexandria, VA. ACM Workshop on privacy in the Electronic Society.

Ibrahim, Y. (2009). Social Networking Sites (SNS) and the'Narcissistic Turn'. The politics of self-exposure. In S. Rummler & K. B. Ng (Eds.), Collaborative Technologies and Applications for Interactive Information Design: Emerging Trends in User Experiences (pp. 82-95.). Hershey, PA: Idea Group Inc.

Keen, A. (2007). The cult of the amateur: How today's Internet is killing our culture. New York: Doubleday/Currency.

Kleinz, T. (2009). Facebook scheitert mit Klage gegen StudiVZ. Retrieved October 27, 2009, from http://www.focus.de/digital/internet/soziale-netzwerke...

Kramer, R.M. (1999). Trust and Distrust in Organizations: Emerging Perspectives, Enduring Questions. Annual Review of Psychology, 50, 569-598.

Kramer, N.C., & Winter, S. (2008). Impression Management 2.0. The relationship of self-esteem, extraversion, self-efficacy, and self-presentation within social networking sites. Journal of Media Psychology, 20(3), 106-116.

Latane, B. (1981). The psychology of social impact. American Psychologist, 36, 343-356.

Lenhart, A. (2009). Adults and social network sites. Retrieved February 17, 2009, from http://www.pewinternet.org/Reports/2009...

Lenhart A., & Madden, M. (2007). Teens, privacy & online social networks. Pew Internet & American Life Project 2007. Retrieved August 8, 2008 from http://www.pewinternet.org...

Lewis, K., Kaufman, J., & Christakis, N. (2008). The taste for privacy: An analysis of college student privacy settings in an online social network. Journal of Computer-Mediated Communication, 14(1), 79-100.

Livingstone, S. (2008). Taking risky opportunities in youthful content creation: teenagers' use of social networking sites for intimacy, privacy and self-expression. New Media & Society, 10, 339-411.

Livingstone, S. (2006). Children’s privacy online: Experimenting with boundaries within and beyond the family. In R. Kraut, M. Brynin, & S. Kiesler (Eds.), Computers, Phones, and the Internet: Domesticating Information Technology (pp. 128–44). Oxford: Oxford University Press.

McKnight, D.H., Choudhury, V., & Kacmar, C. (2002). Developing and validating trust measures for e-Commerce: an integrative typology. Information Systems Research, 13, 334-361.

Orlet, C. (2007, March 2). The look-at-me generation. The American Spectator. Retrieved August 8, 2008 from www.spectator.org/dsp_article.asp?art_id=11093

Pedersen, E.R., LaBrie, J.W., & Lac, A. (2008). Assessment of perceived and actual alcohol norms in varying contexts: Exploring Social Impact Theory among college students. Addictive Behaviors, 33, 552-564.

Raskin, R., & Terry H. (1988). A principal components analysis of the Narcissistic Personality Inventory and further evidence of its construct validity. Journal of Personality and Social Psychology, 54, 890-902.

Rosen, C. (2007). Virtual friendship and the new narcissism. The New Atlantis: A Journal of Technology and Society, Summer, 17, 15-31.

Siibak, A. (2009). Constructing the Self through the Photo selection-Visual Impression Management on Social Networking Websites. Cyberpsychology: Journal of Psychological Research on Cyberspace, 3(1). Retrieved October 27, 2009, from http://cyberpsychology.eu/...

Stein, L., & Sinha, N. (2002). New Global Media and Communication Policy: The Role of the State in Twenty-first Century. In L. Lievrouw & S. Livingstone (Eds.), Handbook of New Media: Social Shaping and Consequences of ICTs (pp. 410–31). London: Sage.

Story, L., & Stone, B. (2007, November 30). Facebook retreats on online tracking. New York Times, Retrieved August 8, 2008 from www.nytimes.com/2007/11/30/technology/30face.html

Synovate (2008, September 3). Netherlands in the lead in social networking. Retrieved September 4, 2008 from http://www.synovate.nl/nieuws/20081009001/news.aspx

Thelwall, M. (2008). Social networks, gender, and friending: An analysis of MySpace member profiles. Journal of the American Society for Information Science and Technology, 59, 1321-1330.

Tufekci, Z. (2008). Can you see me now? Audience and disclosure regulation in online social network sites. Bulletin of Science, Technology & Society, 28, 20-36.

Utz S. (2008). (Selbst)marketing auf Hyves [(Self)marketing on Hyves]. In P. Alpar & S. Blaschke (Eds.), Web2.0. Eine empirische Bestandsaufnahme [Web2.0 - an empirical inventory] (pp. 233-258). Wiesbaden, Germany: Vieweg & Teubner.

Van Lange, P.A.M, Van Vugt, M., Meertens, R.M., & Ruiter, R.A.C. (1998). A social dilemma analysis of commuting preferences: The roles of social value orientation and trust. Journal of Applied Social Psychology, 28, 796-820.

Zhao, S., Grasmuck, S., & Martin, J. (2008). Identity construction on Facebook: Digital empowerment in anchored relationships. Computers in Human Behavior, 24, 1816–1836.

Correspondence to:
Sonja Utz
VU University Amsterdam
Department of Communication Science
De Boelelaan 1081
1081HV Amsterdam
The Netherlands
Phone: +31 20 5989184
Fax: +31 20 5983733
E-mail: s.utz(at)fsw.vu.nl

About author(s)

Author photoSonja Utz is an assistant professor at the Department of Communication Science, VU University Amsterdam. She graduated in psychology at the Catholic University of Eichstatt (Germany) where she also received her Ph.D. in 1999. Her research interests include social processes in social network sites and other virtual communities, social dilemmas in cyberspace, trust in cyberspace (e.g. eBay), and knowledge management.
Author photoNicole Kramer is Professor for "Social Psychology - Media and Communication" at the University Duisburg-Essen since 2007. She finished her PhD in Psychology at the University of Cologne in 2001. In the academic year 2003/2004 she was visiting scholar and visiting lecturer at the University of Cambridge, Faculty of Social and Political Sciences. In 2006 she received the venia legendi for psychology with a habilitation thesis on "Social effects of embodied conversational agents" at the University of Cologne. Her research interests include human-computer-interaction, social psychological aspects of Web 2.0, nonverbal behaviour and computer supported instructional communication.