Convenient or scary? : Understanding the effects of different levels of personalised advertising on message attitudes and disclosure intentions.

Graduate School of Communication

Faculty of Social and Behavioural Sciences Dr. Nadine Bol

Convenient or scary?

Understanding the effects of different levels of personalised advertising on

message attitudes and disclosure intentions.

Master Thesis

Abstract

Due to an increasing accessibility of individual user data, personalised advertising has become omnipresent in daily life. Nevertheless, previous research has shown that in some cases personalisation increased message attitudes and disclosure intentions, whereas in other cases it had a negative impact. To provide a possible explanation for these inconsistent findings and to gain deeper insights into the effects of personalised advertising, this study investigated the impact of low, moderate and high levels of personalisation on message attitudes and disclosure intentions. For this purpose, perceived personalisation, trust in the advertiser and a cost-benefit trade-off (weighing perceived costs against benefits) based on the personalisation-privacy paradox (Awad & Krishnan, 2006) and the privacy calculus theory (Culnan & Armstrong, 1999) were tested as mediators. Additionally, involvement with the advertised product was implemented as a moderator. An online experiment conveyed that personalisation does not necessarily lead to perceived personalisation but showed that perceived personalisation led to higher levels of trust, which led to consumers perceiving more benefits and less costs, which consequently led to more favourable message attitudes and disclosure intentions. However, product involvement only showed to have a direct, positive effect on message attitudes.

Keywords: personalisation, perceived personalisation, trust, personalisation-privacy paradox, message attitudes, disclosure intentions

Convenient or scary? Understanding the effects of different levels of personalised advertising on message attitudes and disclosure intentions.

The continuous development of online technologies like social network sites makes it increasingly easy for advertisers to access and capture personal user data in real time (Couldry & Turow, 2014). The Chinese social media application (app) WeChat, for instance, is installed on nearly every mobile phone in China. It offers a broad variety of features ranging from messaging services, calls, payments and games to options to share content and to connect with nearby strangers. By uniting many features and making other apps redundant, such an app simplifies consumers’ daily lives and is therefore very convenient (Tu, 2016). At the same time, an extensive amount of data is collected through such apps (ibid., 2016). This raises the question whether it is scary that an app gathers so much personal information about consumers.

Given the omnipresence of advertising and mobile devices in today’s society (Chen, Cheng, Yu, & Ju, 2014), consumers increasingly show avoidance reactions (Kitchen & Burgmann, 2011). Research has found that the majority of consumers are sceptical about advertising. Particularly personalised advertising, which can be defined as “advertising that is created for an individual using information about the individual” (Yu & Cude, 2009, p. 504), is often perceived to be obtrusive (Roderick, 2017). Consumers feel that by advertisers using their personal data, they invade their privacy (Baek & Morimoto, 2012). This can lead to feelings of vulnerability (Čaić, Mahr, Aguirre, de Ruyter, & Wetzels, 2015) or annoyance (Bol et al., 2018). At the same time, consumers approve of receiving more relevant (De Keyzer, Dens, & De Pelsmacker, 2015) or appealing (Kim & Han, 2014) messages that can help to save time (Lee & Cranage, 2011). With the emergence of personalised mobile advertising, consumers are continuously exposed to advertising messages promoting products based on their characteristics, behaviour or interests (Chen & Hsieh, 2012). It is therefore not

surprising that consumers become increasingly aware of the costs that come with the advantages of those technologies. The interplay of personalisation, costs and benefits is described in a construct referred to as the personalisation-privacy paradox (Awad & Krishnan, 2006). It depicts a phenomenon that consumers are confronted with when using personalised services: to profit from benefits, they have to accept certain costs that regularly come with personalised services (Lee & Rha, 2016).

To provide personalised advertising and to enable online transactions, it is essential that consumers are sharing data online and approve of their data being used (Dinev & Hart, 2006). The individual’s willingness to provide personal information can also be referred to as disclosure intentions (Olivero & Lunt, 2004). Since consumers form intentions based on attitudes (Mitchell & Olson, 1981), the study will further investigate these. Message attitudes refer to the consumers’ overall evaluation of the advertising message (Belch & Belch, 2015). On the one hand, personalisation has been shown to evoke negative attitudes towards the message (Malheiros, Jennett, Patel, Brostoff, & Sasse, 2012) and to decrease the consumers’ willingness to disclose personal information (Awad & Krishnan, 2006). On the other hand, personalisation has also been shown to positively influence these outcomes by increasing the message’s relevance (Li & Liu, 2017; Maslowska, Smit, & Van den Putte, 2016; Pauzer, 2016; Walrave, Poels, Antheunis, Van den Broeck, & Van Noort, 2016; De Keyzer et al., 2015). Due to potential message avoidance through personalisation and in particular because of the ambiguous if not contradictory results (Li & Liu, 2017), further investigations are inevitable.

To better understand the impact of personalised advertising on consumers’ message attitudes and disclosure intentions, it is first of all essential to understand how consumers form attitudes and intentions. According to the theory of planned behaviour (Ajzen, 1991), consumers hold certain beliefs that then translate into certain attitudes and intentions. The

ambiguous results created by personalised advertising might therefore be justified by the complex trade-off of costs (i.e., messages’ obtrusiveness) and benefits (i.e., messages’ relevance), which are equitable with the beliefs in Ajzen’s theory. According to the personalisation-privacy paradox (Awad & Krishnan, 2006), consumers are willing to provide information as long as the level of gratification gained exceeds the level of risk taken. Furthermore, the so-called privacy calculus theory (Culnan & Armstrong, 1999) suggests that trust in the advertiser counterbalances privacy concerns and thereby positively influences disclosure intentions. Trust refers to the individual’s “willingness to rely on an exchange partner in whom one has confidence” (Moorman, Zaltman, & Deshpandé, 1992, p. 315). By combining these approaches, it can be argued that the synergy of trust, the perceived costs and benefits determine whether the message attitudes turn out rather positive or negative, which in turn influences the consumers’ intention to disclose personal data. Furthermore, it is expected that not only the source of the message but also the consumers’ involvement with the advertised product has an impact on the effects of personalisation on attitudes and disclosure intentions (Lee & Rha, 2016). Product involvement describes the consumers’ relation to the advertised product; more specifically, the “[…] arousal, interest, or motivation for a specific product” (Li & Liu, 2017, p. 134). In line with the logic of dual process models (i.e., the elaboration likelihood model (ELM); Petty & Cacioppo, 1986), involvement is expected to strengthen the effects of personalisation on the cost-benefit trade-off by processing the advertising message more deeply. Consequently, this is expected to influence message attitudes and disclosure intentions. However, Li (2016) found that it is not the personalisation itself that best predicts the effects of personalisation on consumers but that it is perceived personalisation, meaning the consumers’ individual understanding of how strong an advertisement is personalised. Nonetheless, many studies showed that personalisation translated into perceived personalisation (e.g., Li & Liu, 2017; Lee & Cranage, 2011). To get

an even more integrated understanding of personalisation processes, both constructs will be entered into the equation.

As the cumulated effects of these factors could explain the different outcomes of previous studies, this paper aimed to investigate how personalisation translates into perceived personalisation, how the perception of personalisation affects trust, how this series of effects impacts the cost-benefit trade-off, and how the resulting trade-off affects the message attitudes and disclosure intentions. Subsequently, it aimed to investigate the impact of product involvement on this series of effects.

RQ: How does personalisation influence perceived personalisation, trust, perceived benefits and costs, consumers’ attitudes towards the message and, subsequently, their disclosure intentions, and what is the moderating effect of involvement with the advertised product?

Theoretical background The effects of personalisation

Personalisation and perceived personalisation. Advertising messages can be personalised by using different types of personal data. These different types of personal data were found to have varying effects on consumers’ message attitudes and disclosure intentions (e.g., Walrave et al., 2016). Moreover, the different types of personal data used to personalise messages led to variations of how strongly personalised messages were perceived to be (Lee & Cranage, 2011). Personalisation is expected to be most effective when consumers are aware of the personalisation (Maslowska et al., 2016). Therefore, different types of personal data can be used to create different levels of (perceived) personalisation. Since previous studies showed that different levels of personalisation translated into respective levels of perceived personalisation (e.g., Li & Liu, 2017), the following is hypothesised.

H1: Low, moderate and high levels of personalisation will translate into respective levels of perceived personalisation.

Personalisation and perceived benefits and costs. When a message is personalised and consumers perceive the message to be personalised, there can be certain costs and benefits resulting from that message. Through feelings of perceived personalisation, personalisation can evoke negative feelings; for instance, privacy concerns or “the degree to which a consumer is worried about the potential invasion of the right to prevent the disclosure of personal information” (Baek & Morimoto, 2012, p. 63). Consequently, individuals sometimes experience feelings of vulnerability (Čaić et al., 2015) or irritation, which can originate when consumers are surprised by the advertiser having access to their personal data (Yang, Kim, & Yoo, 2013). These mechanisms are summarised as perceived costs. High levels of personalisation are expected to produce stronger perceptions of costs than low or moderate levels of personalisation. However, disclosing personal data for advertising purposes can also lead to certain advantages. Perceived benefits can be the extent to which the advertising message was seen to be more relevant, useful (De Keyzer et al., 2015), informative (Kim & Han, 2014, p. 257), increased the consumers’ productivity (Lee & Cranage, 2011) or offered entertainment (Kim & Han, 2014). Same as for the costs, it is expected that high levels of personalisation offer more benefits, ergo stronger perceptions of benefits than low or moderate levels of personalisation. This leads to the following hypothesis.

H2: High levels of personalisation compared to low and moderate levels of personalisation, as well as moderate compared to low levels of personalisation, will increase perceived personalisation and will consequently enhance (a) perceived costs and (b) perceived benefits of disclosing personal data.

Personalisation and the formation of attitudes and intentions. Applying the theory of planned behaviour (Ajzen, 1991) to the context of personalised advertising, trust, costs and benefits associated with disclosing personal data constitute the basis for forming attitudes and subsequently intentions (Cheung & To, 2017; Ajzen, 1991).

The cost-benefit trade-off. When consumers hold both negative and positive beliefs it brings them into a dilemma when forming attitudes and intentions. Negative beliefs are expected to lead to unfavourable attitudes and intentions whereas positive beliefs are expected to lead to favourable attitudes and intentions. When consumers hold both beliefs it is questionable what the overall outcomes will be. This dilemma has been expressed by the personalisation-privacy paradox (Awad & Krishnan, 2006). It is referred to as a paradox because consumers paradoxically wish for personalisation and the corresponding benefits without wanting to bear the related costs (ibid., 2006). Therefore, the paradox suggests that consumers are willing to provide information as long as the level of gratification gained exceeds the level of risk taken. Thus, if consumers feel that the advertising message gives them a great advantage, they are more likely to accept a higher risk and vice versa when sharing personal data (Lee & Rha, 2016).

As previous studies showed that perceptions of risk and disclosure intentions were only slightly related (Baruh, Secinti, & Cemalcilar, 2017), a subsequent concept will be added to explain the cost-benefit trade-off in regard to attitudes and intentions. The privacy calculus theory (Culnan & Armstrong, 1999) argues that consumers weigh risks, such as privacy concerns, less heavy when they trust the advertiser and are therefore more willing to disclose their personal data. Supplementary, it assumes that perceived benefits often carry more weight than perceived cost. This weighting can be explained by the fact that perceived benefits offer an immediate gratification to the consumer, whereas effects caused by the perceived costs often lie in the future. As most people are subject to self-control problems,

immediate effects carry more weight than long-term consequences (Acquisti & Grossklags, 2005).

In line with these assumptions, it is expected that the perceived benefits outweigh the perceived costs for moderate levels of personalisation and perceived personalisation. However, it is expected that costs will increase more strongly than benefits. Therefore, low levels are expected to evoke little benefits and costs, leading to a neutral cost-benefit trade-off and high levels are expected to evoke the most benefits but even more costs (e.g., Čaić et al., 2015; Malheiros et al., 2012). Thus, perceived costs are expected to outweigh perceived benefits of a highly personalised message. With moderate levels leading to stronger perceptions than low levels of personalisation and high levels of personalisation evoking strong negative feelings, moderate levels are expected to be most effective in regard to finding the balance between making the advertisement noticed, interesting and comfortable at the same time (Malheiros et al., 2012). To provide clarification for the cost-benefit trade-off evoked by personalisation and perceived personalisation, the following is hypothesised.

H3: High levels of personalisation compared to low and moderate levels of personalisation, as well as moderate compared to low levels of personalisation, will increase perceived personalisation. Consequently, high levels of perceived personalisation compared to low and moderate levels of perceived personalisation will negatively influence the cost-benefit trade-off whereas moderate compared to low levels of perceived personalisation will positively influence the cost-benefit trade-off.

The mediating effect of trust. Since online settings offer little cues for consumers to base their decisions on, trust is expected to substitute these by reducing complexity (Bleier & Eisenbeiss, 2015). Personalisation and perceived personalisation are expected to enhance trust in the advertiser because consumers feel to be known by the advertiser, creating a more

familiar atmosphere (Walrave et al., 2016). In line with the privacy calculus theory (Dinev & Hart, 2006; Culnan & Armstrong, 1999), trust is expected to give consumers a feeling of safety, which consequently reduces probable privacy concerns (Bleier & Eisenbeiss, 2015). According to dual process models such as the heuristic-systematic model (Chen & Chaiken, 1999) and the ELM (Petty & Cacioppo, 1986), consumers have two different possibilities of processing information – centrally or peripherally. By taking the central route, consumers focus on arguments and deeply process information; in contrast, taking the peripheral route, consumers rely on cues, processing information less deeply. Hence, a high level of trust can serve as a cue, making the consumers process the advertising message less deeply and thereby evoking less cost (Walrave et al., 2016) and more benefit perceptions (Cheung & To, 2017; Beldad, De Jong, & Steehouder, 2011). Thus, the cost-benefit trade-off will be balanced in a more favourable way: the strong negative feelings evoked by high levels of personalisation and perceived personalisation are expected to be lessened. This results in benefits exceeding costs. With high levels of personalisation leading to stronger perceptions than moderate and low levels of personalisation, and moderate levels leading to stronger perceptions than low levels of personalisation, high levels are expected to evoke the highest perceptions of personalisation and trust and consequently the most favourable cost-benefit trade-off. Therefore, the following is hypothesised.

H4: High levels of personalisation compared to low and moderate levels of personalisation, as well as moderate compared to low levels of personalisation, will increase perceived personalisation, which will increase (a) trust, which consequently will positively influence the (b) cost-benefit trade-off.

Message attitudes and disclosure intentions. Since the cumulated effects of personalisation on perceived personalisation, trust and the cost-benefit trade-off are expected

to constitute the basis for the formation of attitudes and intentions, these outcomes are expected to be influenced accordingly (Cheung & To, 2017; Ajzen, 1991). Thus, message attitudes and disclosure intentions are expected to be the higher, the higher the level of personalisation, perceived personalisation, trust, and therefore, the more favourable the cost-benefit trade-off. This leads to the following hypothesis.

H5: High levels of personalisation compared to low and moderate levels of personalisation, as well as moderate compared to low levels of personalisation, will increase perceived personalisation, which will increase trust. This will enhance the cost-benefit trade-off, which consequently will positively impact (a) message attitudes and (b) disclosure intentions.

The moderating role of involvement

As involvement has demonstrated to strengthen the effects of personalisation in previous research (Li & Liu, 2017; Lee & Rha, 2016; De Keyzer et al., 2015), this factor could further broaden the understanding of different outcomes of previous studies and the cost-benefit trade-off. Conversely to the facilitating effect of trust, involvement is expected to cause a deeper processing of the advertising message and to thereby strengthen both the beneficial and costly aspects evoked by personalisation and perceived personalisation (Lee & Rha, 2016). The ELM states that the motivation to process depicts one of the key prerequisites for deeply processing information and that motivation is higher when involvement is high (Petty & Cacioppo, 1986). Therefore, when personalisation and perceived personalisation is interacting with deeper processing evoked by high involvement with the product, both the positive, e.g., relevance (De Keyzer et al., 2015), and negative aspects, e.g., discomfort (Lee & Rha, 2016) are expected to increase. Thus, under high levels of involvement compared to low levels of involvement, perceived benefits are expected to

increase for all levels of personalisation and perceived personalisation. As there are little costs that could be strengthened for low levels of personalisation and perceived personalisation, the cost-benefit trade-off and subsequently message attitudes and disclosure intentions are expected to be positively influenced under high levels of involvement compared to low levels of involvement. Due to higher costs in the remaining levels of personalisation and perceived personalisation, the cost-benefit trade-off and subsequently the message attitudes and disclosure intentions are not expected to be influenced by involvement. Thus, the following hypothesis is posed.

H6: Under high levels of involvement compared to low levels of involvement, low levels of personalisation inducing lower levels of perceived personalisation will result in a more positive (a) cost-benefit trade-off compared to moderate and high levels of personalisation. Consequently, this will lead to more favourable (b) message attitudes and (c) disclosure intentions.

The following model summarises the hypotheses and effects described above.

Figure 1. Conceptual model.

Methods Power analysis

A power analysis using the software G*Power was conducted to ensure sufficient power for hypotheses-testing (Faul, Erdfelder, Buchner, & Lang, 2009). The analysis was based on a medium effect size (f = .25). The analysis revealed that to achieve a power of 80% with an alpha level of 5% (two-tailed), a minimum sample size of N = 211 is required. The final number of participants (N = 376) resulted in an achieved power of 98%.

Participants

A convenience sample, mainly consisting of university students, was recruited via social media. Of 480 initial participants, 104 were excluded as they had no smart device, did not complete the essential survey questions or finished the survey in under three minutes, which was the minimum of time needed to complete the survey. All remaining participants gave their informed consent to participate in the study, leading to a final sample size of N = 376. The final sample was aged between 18 and 72 (M = 31.81, SD = 12.07) and mainly consisted of females (69%). The majority of participants were German (69%) or Dutch (14%). With regard to education, most participants achieved either a Bachelor's (48%) or Master’s

degree (26%).

16% graduated from high school, 5% completed an apprenticeship, 4% were doctorates and only 1% achieved less than a high school degree.

Experimental design and procedure

A scenario-based online experiment consisting of a 3 × 2 between-subjects factorial design with three levels of personalisation (namely: low, moderate and high) and two levels of involvement (namely: low and high) was performed. Participants started the online survey by giving their consent to participate, filling out their demographical data and indicating whether they had a smartphone. Subsequently, they were asked to imagine that they were looking at the news feed of a social media app on their smart device, displaying an advertisement. The different levels of personalisation were created by using different types of personal data. The

types of personal data and the involvement with different products were based on a pilot study (see Appendix A). Accordingly, the participants’ gender was implemented for the low personalisation condition, their location for the moderate and their credit card number for the high personalisation condition. Hence, the advertisement gave the impression that these data were known by the advertiser. Regarding involvement, the participants were either exposed to an advertisement for painkillers (low involvement) or travels (high involvement). Each participant was randomly assigned to one of six experimental conditions (see Appendix B). After the exposure to the stimulus, the level of perceived personalisation was assessed and the level of involvement with the product was checked for manipulation. Subsequently, the level of trust, the participants’ benefit and risk perception, their message attitudes as well as their disclosure intentions were measured. Towards the end of the survey, participants were asked how realistic they perceived the advertisement to evaluate the meaningfulness of the manipulation. Lastly, participants were debriefed to reinsure that they knew the scenario was fictitious and their data had not been misused or saved.

Measures

An extensive overview over all items used to measure these concepts is given in Table C1 (see Appendix). As consumers cannot choose to avoid specific costs or the disclosure of specific data, all items of each scale were included as one factor, even if factor analysis suggested that there was more than one dimension presented by the scale. As all factors were loaded above .45 (see Table 1), this was also a statistically feasible choice. The table further indicates the level of reliability and the distribution of the scales. Unless indicated differently, all items were rated on a 7-point Likert scale anchored by 1 (strongly disagree) and 7 (strongly agree).

Perceived personalisation. Perceived personalisation (M = 3.00, SD = 1.30) measured to what extent participants felt that the advertisement was individually created for

them. It was measured with five items (Song, Kim, Kim, Lee, & Lee, 2016; Kalyanaraman & Sundar, 2006). One of the items read: “The advertisement directly addressed me as an individual”. To fully exhaust the collected data, the perceived level of personalisation was included as a mediator to obtain a deeper understanding of personalisation effects rather than merely detecting whether there were significant differences between the experimental conditions by conducting a manipulation check (O'Keefe, 2003).

Table 1

Factor analyses, reliability tests and distribution

Variable Factor Eigen-

value % of variance Cumu- lative % Cron-bach’s α Skewness (SE = .13) Kurtosis (SE) perceived personalisation involvement trust perceived benefits perceived costs message attitude disclosure intention 1 1 1 1 1 1 1 3.02 6.13 3.24 5.15 6.43 5.36 5.51 60.41 87.53 81.02 64.33 49.48 66.96 50.11 60.41 87.53 81.02 64.33 49.48 66.96 50.11 .82 .98 .92 .92 .94 .93 .90 0.28 0.04 0.13 -0.16 -0.49 0.27 0.42 -0.79 (.25) -1.52 (.25) -0.68 (.25) -0.77 (.25) -0.27 (.27) -0.64 (.25) -0.52 (.26)

Involvement. Involvement (M = 3.91, SD = 2.11) described consumers’ interest in the advertised product. Determining the success of the involvement manipulation was accomplished by asking the participants how well the seven items (Hong, 2015), e.g., “[The product] is important to me”, applied to the presented product. The items were based on Zaichkowsky’s (1994) revised personal involvement inventory.

Trust. Trust (M = 3.13, SD = 1.29) referred to the extent consumers rely on the advertiser and was measured with four items (Bol et al., 2018). One example item was: “I trust that [the advertiser] handles my personal information correctly”.

Perceived costs. Perceived costs (M = 5.52, SD = 0.94) were the disadvantages that the consumers experienced as a result of disclosing personal data. The measurement used five items introduced by Bol et al. (2018) and three items introduced by Kim & Han (2014). For example, one item read: “Personal information (such as my online search behaviour, location, age and gender) being used for online advertising is risky” (Bol et al., 2018). Moreover, five items assessed privacy concerns, such as “I am concerned that when I visit a social media app as described in the scenario, my personal information (such as my online search behaviour, location, age and gender) can be misused by others” (ibid., 2018). The item “Personal information […] being used for online advertising is safe” (ibid., 2018) was recoded into a new variable because it needed to be reversely coded to match the other items.

Perceived benefits. Perceived benefits (M = 3.58, SD = 1.32) were the profits that the consumer felt to extract from disclosing personal data. The measurement used eight items (Bol et al., 2018), such as “Personal information (such as my online search behaviour, location, age and gender) being used for online advertising makes me receive relevant suggestions”.

Cost-benefit trade-off. The cost-benefit trade-off (M = -2.01, SD = 1.88) was created by subtracting the perceived costs from the perceived benefits, resulting in a scale ranging from -6.00 to 2.77. Therefore, perceived costs exceeded perceived benefits for all values below zero.

Message attitude. The message attitude (M = 2.96, SD = 1.25) described the participants’ overall evaluation of the advertisement and was measured on eight items (Maslowska et al., 2016). One example item was: “The advertisement I just saw is attractive”.

Disclosure intention. The consumers’ disclosure intention (M = 2.61, SD = 1.17) referred to the consumers’ willingness to share personal information online, and was assessed by eleven items (Bol et al., 2018). The questions stated: “Imagine you were to visit a comparable social media app again in the future. How likely is it that you would share the following personal information with this app to receive personalized advertisements?” Participants rated the items (e.g., “Your gender”) on a 7-point Likert scale anchored by 1 (very unlikely) and 7 (very likely).

Control variables. Of several control variables measured in the survey, namely, reality of the scenario, nationality, level of education, gender and age, only the latter significantly influenced the outcome variables (see Appendix D) and was therefore included as a control variable.

Statistical analyses

The hypotheses were tested using SPSS through an analysis of covariance (ANCOVA) and mediation-moderation analyses using Hayes’ PROCESS macro (Hayes, 2017). All analyses were examined with age as control variable. The ANCOVA was conducted to test H1 with personalisation as independent variable and perceived personalisation as dependent variable. H2 was tested by mediation analyses using Hayes’ PROCESS Model 4 with personalisation as independent variable, perceived personalisation as mediating variable, and perceived benefits and costs as dependent variables. The analyses were repeated three times to test all levels of personalisation against each other. In each case, one of the dummy-coded levels of personalisation served as independent variable and another one as control variable, making the excluded dummy-coded level of personalisation the reference group. This approach was used for all subsequent mediation analyses. To test H3-H5, serial mediation analyses using Hayes’ PROCESS Model 6 with personalisation as independent variable,

perceived personalisation, trust, the cost-benefit trade-off and message attitudes as mediating variables, and disclosure intention as dependent variable. A moderated serial mediation analysis using Hayes’ PROCESS Model 92 was conducted with personalisation as independent variable, perceived personalisation, trust, the cost-benefit trade-off and message attitudes as mediating variables, disclosure intention as dependent variable and involvement as moderating variable to test H6.

Results Randomisation check

A one-way analysis of variance (ANOVA) revealed that the participants’ age did not significantly differ across experimental conditions, F(5, 373) = 0.23, p = .949, ηp2 = .00. Furthermore, a chi-square test of independence showed that neither the participants’ gender, 2_{(5) = 6.13, p = .294, education, }2_{(25) = 18.15, p = .836, nor nationality, }2

(150) = 154.65, p = .381, differed significantly across the experimental conditions, showing successful randomisation. However, an ANOVA revealed that the reality of the scenario significantly differed between the levels of personalisation, F(2, 368) = 6.23, p = .002, ηp2 = .03. Post-hoc tests showed that for high levels of personalisation (credit card number; M = 3.60, SD = 1.75), participants rated reality of the scenario significantly lower than for moderate (location; M = 4.18, SD = 1.62, p = .029) or low (gender; M = 4.35, SD = 1.82, p = .002) levels of personalisation. Moderate and low levels of personalisation did not significantly differ in the participants’ evaluation of the reality of the scenario, p = 1.000. Nevertheless, of these potential control variables, only age significantly influenced the outcome variables (see Appendix D), and was therefore included as a control variable in subsequent analyses.

Manipulation check

The manipulation of product involvement was successful. As expected, an independent samples t-test showed that travels (M = 5.76, SD = 1.10) involved the participants significantly

stronger than painkillers (M = 2.12, SD = 1.04), t(373) = 32.92, p < .001, 95% CI [3.85, -3.42].

Main analysis

H1: perceived personalisation. Personalisation had no significant effect on perceived personalisation (see Figure 2), F(2, 367) = 0.02, p = .976, ηp2 = .00. Low levels of personalisation did not lead to significantly lower levels of perceived personalisation (M = 2.90, SD = 1.29), nor did moderate (M = 2.96, SD = 1.26) and high levels of personalisation (M = 2.97, SD = 1.34) translate into the respective levels of perceived personalisation. Therefore, H1 was rejected.

H2: perceived benefits and costs. The effect of personalisation on perceived costs, e.g., b = -0.01, SE = .02, 95% CI [-0.045, 0.037], and benefits, e.g., b = 0.01, SE = .07, 95% CI [-0.125, 0.136], was not mediated by perceived personalisation due to personalisation and perceived personalisation not being correlated, e.g., b = 0.1, SE = .18, 95% CI [-0.250, 0.447] (see Table D1 and Figures D1&D2 (Appendix)). Therefore, H2 was rejected. Even though H2 was rejected, perceived personalisation was negatively correlated with perceived costs, b = -0.11, SE = .04, 95% CI [-0.185, -0.026], suggesting that higher levels of perceived personalisation led to significantly lower perceived costs. Furthermore, perceived personalisation was positively correlated with perceived benefits, b = 0.39, SE = .05, 95% CI [0.289, 0.482], meaning that higher levels of perceived personalisation led to higher levels of perceived benefits.

Figure 2. The effect of personalisation on perceived personalisation. Perceived personalisation was rated on a 7-point Likert scale with 1 as lowest and 7 as highest value.

H3: cost-benefit trade-off. The effect of personalisation on the cost-benefit trade-off was not mediated by perceived personalisation, see Table D2 and Figure D3 (Appendix). Therefore, H3 was rejected. Even though H3 was rejected, perceived personalisation was positively correlated with the cost-benefit trade-off, b = 0.33, SE = .08, 95% CI [0.186, 0.482], suggesting that higher levels of perceived personalisation led to a more favourable cost-benefit trade-off. Nonetheless, the cost-benefit trade-off constantly remained negative as a result of perceived costs exceeding benefits, see Figure 3.

1 2 3 personalisation p er ce iv ed p er so n alis atio n

low moderate high

1 4 7 low high m ea n p er ce p tio n o f co sts an d b en ef its perceived personalisation perceived costs perceived benefits

Figure 3. The effects of perceived personalisation on perceived costs and benefits. Perceived costs and benefits were rated on a 7-point Likert scale with 1 as lowest and 7 as highest value.

H4: trust and cost-benefit trade-off. The effect of personalisation on the cost-benefit trade-off was not mediated by perceived personalisation and trust, see Table D2 and Figure D3 (Appendix). Therefore, H4 was rejected. Even though H4 was rejected, mediation analyses showed a significant indirect effect of perceived personalisation on the cost-benefit trade-off via trust, b = 0.14, SE = .04, 95% CI [0.071, 0.212], suggesting that higher levels of perceived personalisation led to higher trust, and higher trust consequently led to a more favourable

cost-benefit trade-off.

H5: message attitudes and disclosure intentions. The effect of personalisation on message attitudes and disclosure intentions was not mediated by perceived personalisation, trust, and the cost-benefit trade-off, see Table D2 and Figure D3 (Appendix). Therefore, H5 was rejected. Even though H5 was rejected, mediation analyses showed a significant indirect effect of perceived personalisation on message attitudes via trust and the cost-benefit trade-off, b = 0.01, SE = .01, 95% CI [0.003, 0.027]. This suggests that higher levels of perceived personalisation led to higher trust, which led to a more favourable cost-benefit trade-off. Consequently, a more favourable cost-benefit trade-off led to more favourable message attitudes. Similarly, mediation analyses showed a significant indirect effect of perceived personalisation on disclosure intentions via trust and the cost-benefit trade-off, b = 0.03, SE = .01, 95% CI [0.014, 0.052]. This implies that higher levels of perceived personalisation led to higher trust, which led to a more favourable cost-benefit trade-off, and consequently more favourable disclosure intentions. However, the effect of perceived personalisation on disclosure intentions, e.g., b = 0.00, SE = .00, 95% CI [-0.000, 0.001], was not mediated by

trust, the cost-benefit trade-off and message attitudes, see Table D2 and Figure D3 (Appendix).

H6: moderating effect of involvement. The moderated serial mediation showed that there were no significant interactions caused by perceived personalisation and product involvement, e.g., b = 0.27, SE = .15, 95% CI [-0.019, 0.560], see Figure 4 & D4 and Table D3 (Appendix). Therefore, H6 was rejected.

Figure 4. The effect of perceived personalisation and involvement on message attitudes. Message attitudes were rated on a 7-point Likert scale with 1 as lowest and 7 as highest value.

Discussion

The study aimed to find out how personalisation influences consumers’ perceived personalisation, trust, perceived costs and benefits, attitudes towards the message and subsequently their disclosure intentions. As, against expectations, the different levels of personalisation did not translate into respective levels of perceived personalisation, the effect of personalisation on message attitudes and disclosure intentions could not be explained by perceived personalisation, trust, and the cost-benefit trade-off. However, a large part of the underlying mechanisms induced by perceived personalisation were explained by the hypothesised model. It showed that perceived personalisation positively influenced trust, perceived costs and benefits, and thereby the cost-benefit trade-off, and consequently

1 2 3 4 low high m ess ag e attitu d es perceived personalisation low involvement high involvement

consumers’ attitudes towards the message and their disclosure intentions. Merely the negative impact of perceived personalisation on perceived costs and consequently the positive impact on the cost-benefit trade-off, message attitudes and disclosure intentions was contrary to the expectations that higher levels of perceived personalisation would induce higher levels of perceived costs. The study further investigated the moderating effect of involvement with the advertised product. Results showed that there were no interaction effects caused by involvement.

The fact that personalisation did not translate into perceived personalisation emphasized the distinctiveness of the two concepts. Since consumers often do not know their preferences, reality does not necessarily translate into the respective perceptions (Simonson, 2005). Thus, statements of what consumers want often differ from what they actually want. This variation can be explained by the fact that personalisation is created by the advertiser and is therefore controlled by the advertiser’s sphere, whereas perceived personalisation lies in the consumers’ sphere (Li, 2016). Lying in the consumers’ sphere, perceived personalisation might create more positive associations than personalisation, which lies in the advertiser’s sphere and is based on the advertiser’s expectations. Thereby, Li’s (2016) findings that perceived personalisation is a better predictor for outcomes than personalisation could be supported. In line with these assumptions, it could be argued that personalisation is a rather instrumental construct that needs to be processed more centrally (De Keyzer et al., 2015) and that it is more closely related to negative aspects (French et al., 2005). This leads to increasing costs when personalisation is high. In contrast, perceived personalisation is rather affective, does not need deeper processing, can therefore rather serve as a cue (De Keyzer et al., 2015) and is more strongly associated with positive aspects (French et al., 2005). This leads to increasing benefits and decreasing costs. Whilst personalisation is rather based on facts, perceived personalisation can represent something different for every individual (O'Keefe,

2003). Thus, perceived personalisation could be based on the consumers’ characteristics and preferences. For instance, a private person might perceive personalised content as more obtrusive than an extrovert person or, in line with the findings regarding the control variable (see Appendix D), an older person being less familiar with new technologies might not identify personalised content as such whereas a young person would.

Another important finding of this study is that, in line with the personalisation-privacy paradox (Awad & Krishnan, 2006) and the privacy calculus theory (Culnan & Armstrong, 1999), consumers seem to base their message attitudes and disclosure intentions on a trade-off of trust, costs and benefits. Due to the most positive cost-benefit trade-off in high levels of perceived personalisation, message attitudes and disclosure intentions were higher for high levels than for low levels of perceived personalisation. Although perceived costs still exceeded perceived benefits, this suggests that consumers’ dilemma of weighing benefits against costs decreases as the level of perceived personalisation increases. When they perceive the message to be highly personalised they perceive less costs and more benefits. Accordingly, their interests become less paradoxically. As the cost-benefit trade-off remained negative for all conditions, message attitudes and disclosure intentions still remained rather unfavourable. Thus, outcomes induced by a positive cost-benefit trade-off can only be assumed to align with the theory. The fact that there was no positive cost-benefit trade-off for any condition can be explained by Tversky & Kahneman’s prospect theory (1981) arguing that losses generally loom larger than gains. This means that the severity of costs will be weighed heavier than the gratification of benefits. Consumers’ general scepticism towards advertising reinforces this phenomenon (Lee & Rha, 2016). Regardless, the argumentation of the privacy calculus theory (Culnan & Armstrong, 1999) that immediate gratification would outweigh long-term risks could not be supported. Instead, the assumption that trust has a positive influence on the cost-benefit trade-off was supported for perceived personalisation.

These findings also reinforce the explanation that trust serves as a cue to peripherally process the advertising message. However, product involvement showed not to influence the cost-benefit trade-off and therefore does not depict a sensible addition to the calculus theories. Limitations

Despite these relevant findings for the field of personalised advertising, some limitations should be addressed. First of all, online surveys usually do not manage to properly involve its participants. This can lead to participants not properly processing and paying little attention. Although a certain involvement with the products was supported by the study, involvement with the survey and study materials could have been rather low. Especially scenario based online experiments do not enable their participants to experience the constructs exactly in the way they would in reality. Therefore, participants usually show lower involvement with the scenarios than they would in reality, leading to weaker reactions. Participants who processed the survey more peripherally might therefore also explain the stronger effects of perceived personalisation compared to personalisation, as perceived personalisation is expected to require less depth in processing. Moreover, effect sizes in communication effects are often small (Kramer, Guillory, & Hancock, 2014). Particularly with personalisation being so commonly used, consumers assimilated to personalised content and show little reaction, ergo perceptions of personalisation. However, considering the large scale of mobile advertising, even very small effects can lead to large results (ibid., 2014). Another limitation can be seen in the uniformity of the sample. The results can be considered as quite reliable concerning the specific consumer group of young, highly educated people but not necessarily for a whole population (Campbell, 1957). For example, people with these characteristics have shown to be less concerned about their privacy (Smit, Van Noort, & Voorveld, 2014). Therefore, repeating the study with a sample that allows results to be more

reflective of the whole population could lead to different results regarding the cost-benefit trade-off and subsequently message attitudes and disclosure intentions.

Future Research

As the model revealed many interesting findings for perceived personalisation (i.e., support for a cost-benefit trade-off), future research should invest in learning more about factors that predict perceived personalisation to create personalised messages that evoke these perceptions and subsequently corresponding message attitudes and disclosure intentions. Previous research suggests that personalisation is not perceived to be personalised either because advertisers are unable to determine consumers’ wants and needs or because they fail to correctly match these wants and needs to the right consumer (e.g., Kramer, Spolter-Weisfeld, & Thakkar, 2007). To improve the former, advertisers could focus more on implicit rather than explicit measurements. As consumers fail to express their wants and needs (Simonson, 2005), implicit measures could be more accurate in learning about their wants and needs (Belch & Belch, 2015). To improve the latter, it could be investigated whether personalisation better predicts perceived personalisation when it interacts with consumers’ characteristics and preferences. For example, consumers could perceive the same personalised message as more or less personalised depending on their knowledge about personalised services. Moreover, Rimer & Kreuter (2006) suggest that personalisation can be achieved by providing information matching consumers’ wants and needs, by providing the message in an interesting context, by the design of the message or by the channels and frequency in which the message is sent. These approaches to personalise advertising messages could have a varying impact on consumers’ perceived personalisation. To investigate how well these different techniques translate into perceived personalization, the so-called goodness-of-fit could be analysed (Kreuter, Oswald, Bull, & Clark, 2000). Kreuter et al. (2000) found that messages were more effective when they were a good fit, i.e. relevant to the customer, than when they were not.

Hence, a goodness-of-fit measurement could be another approach to clarify the relationship of personalisation and perceived personalisation.

Implications for Research and Practice

The findings of this study suggest some implications for both research and practice. Since personalisation and perceived personalisation seem to trigger different processes, these two distinctive constructs should be treated correspondently. Other than for high levels of personalisation, high levels of perceived personalisation seem to bear no risk of creating costs. Therefore, the more a message is perceived as personalised, the better. To create advertising messages that evoke favourable attitudes and disclosure intentions, advertisers need to personalise messages in such a way that consumers are aware of them being personalised (Li, 2016). For example, the data used to manipulate personalisation in this study might have rather evoked different levels of data sensitivity than perceived personalisation. Thus, it should first be investigated how the message is best created to evoke perceived personalisation. This could be achieved by measuring the goodness-of-fit of the approaches to personalise a message with different types of consumers. Beyond, ensuring high levels of perceived personalisation is highly advisable. These showed to reduce the danger of yielding negative feelings, leading to more favourable attitudes and disclosure intentions. This in turn possibly lessens the consumers’ avoidance reactions to advertising messages.

Moreover, future research can learn from this study that it is essential to carefully align the concepts tested. For instance, the fact that attitudes did not predict disclosure intentions could be explained by the fact that the study measured message attitudes and not disclosure attitudes. As message attitudes and disclosure attitudes are two different constructs, no prediction of disclosure intentions could be made from message attitudes. To provide support for the theory of planned behaviour (Ajzen, 1991), the study would have to be repeated with a measure of disclosure attitudes. Similarly, the lack of effect that product involvement had on

perceived benefits, costs and disclosure intentions could be due to the fact that the study measured disclosure benefits, costs and intentions. Although product involvement showed to be an effective moderator in previous studies (e.g., Li & Liu, 2017), it did not yield significant results in this study. Possibly, disclosure involvement could rather cause significant results and thereby depict an interesting subject for further research.

Another implication follows from the findings about the personalisation-privacy paradox (Awad & Krishnan, 2006) and the privacy calculus theory (Culnan & Armstrong, 1999). An effective advertising message balances trust, perceived costs and benefits in such a way, that they evoke positive message attitudes and disclosure intentions. However, for the cost-benefit trade-off perceived costs have shown to weigh heavier than perceived benefits. Among others, a positive cost-benefit trade-off could be achieved by trust-building strategies. Therefore, trust in the advertiser could be a sensible addition to the personalisation-privacy paradox (Awad & Krishnan, 2006).

Conclusion

This study demonstrated to explain a large part of the underlying mechanisms induced by perceived personalisation. It clarifies the effect of perceived personalisation through a deeper understanding of the personalisation-privacy paradox (Awad & Krishnan, 2006) and the privacy calculus theory (Culnan & Armstrong, 1999) and finally gives some indication of factors influencing this indirect effect. When future research manages to provide the link between personalisation in advertising messages and perceived personalisation, this study eminently contributes to the understanding of the effects of personalised advertising. But even after fully understanding processes evoked by personalisation and perceived personalisation, advertisers’ possibilities to personalise advertising should be regulated. Particularly because the omnipresence of personalised advertising makes it nearly impossible for consumers to opt out without partly excluding themselves from society (e.g., WeChat).

Personalised content based on advertiser’s expectations or research might still lead to a personalisation bias, which might neglect consumers’ actual wants and needs and divide them into groups being exposed to different content. Being exposed to different content can lead to discrimination of certain consumer groups. Hence, advertisers should be forced to acknowledge a certain responsibility when making use of personal data and to act in accordance. Since technologies in this field develop rapidly, administration of justice remains challenged to supervise advertisers’ actions.

References

Acquisti, A. & Grossklags, J. (2005). Privacy and rationality in individual decision making. IEEE Security & Privacy, 3(1), 26-33. doi:10.1109/MSP.2005.22.

Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50(2), 179-211. doi:10.1016/0749-5978(91)90020-T.

Awad, N. & Krishnan, M. (2006). The personalization privacy paradox: An empirical evaluation of information transparency and the willingness to be profiled online for personalization. MIS Quarterly, 30(1), 13-28. Retrieved from http://www.jstor.org/ stable/25148715. doi:10.2307/25148715.

Baek, T. H. & Morimoto, M. (2012). Stay away from me: Examining the determinants of consumer avoidance of personalized advertising. Journal of Advertising, 41(1), 59-76. doi:10.2753/JOA0091-3367410105.

Baruh, L., Secinti, E., & Cemalcilar, Z. (2017). Online privacy concerns and privacy management: a meta‐analytical review. Journal of Communication, 67(1), 26-53. doi:10.1111/jcom.12276.

Belch, G. E. & Belch, M. A. (2015). Advertising and promotion. An integrated marketing communications perspective. New York, NY: McGraw-Hill Education.

Beldad, A., De Jong, M., & Steehouder, M. (2011). I trust not therefore it must be risky: Determinants of the perceived risks of disclosing personal data for e-government transactions. Computers in Human Behavior, 27(6), 2233-2242. doi:10.1016/j.chb. 2011.07.002.

Bleier, A. & Eisenbeiss, 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. Bol, N., Dienlin, T., Kruikemeier, S., Sax, M., Boerman, S. C., Strycharz, J., … De Vreese,

Analyzing self-disclosure across health, news, and commerce contexts. Manuscript in preparation.

Čaić, M. M., Mahr, D., Aguirre, M. E., de Ruyter, M. K., & Wetzels, M. (2015). “Too close for comfort”: The negative effects of location-based advertising. In Advances in Advertising Research, 5, 103-111. doi:10.1007/978-3-658-08132-4_8.

Campbell, D. T. (1957). Factors relevant to the validity of experiments in social settings. Psychological Bulletin, 54(4), 297-312. doi:10.1037/h0040950.

Chen, P. T., Cheng, J. Z., Yu, Y. W., & Ju, P. H. (2014). Mobile advertising setting analysis and its strategic implications. Technology in Society, 39, 129-141. doi:10.1016/j. techsoc.2014.09.004.

Chen, P. T.& Hsieh, H. P. (2012). Personalized mobile advertising: Its key attributes, trends, and social impact. Technological Forecasting and Social Change, 79(3), 543-557. doi:10.1016/j.techfore.2011.08.011.

Chen, S., & Chaiken, S. (1999). The heuristic-systematic model in its broader context. In S. Chaiken & Y. Trope (Eds.), Dual-process theories in social psychology (pp. 73-96). New York: Guilford Press.

Cheung, M. F. & To, W. M. (2017). The influence of the propensity to trust on mobile users' attitudes toward in-app advertisements: An extension of the theory of planned behavior. Computers in Human Behavior, 76, 102-111. doi:10.1016/j.chb.2017.07. 011.

Couldry, N.& Turow, J. (2014). Advertising, big data, and the clearance of the public realm: Marketers' new approaches to the content subsidy. International Journal of Communication, 8, 1710-1726.

Culnan, M. J. & Armstrong, P. K. (1999). Information privacy concerns, procedural fairness, and impersonal trust: An empirical investigation. Organization Science, 10(1), 104-115. doi:10.1287/orsc.10.1.104.

De Keyzer, F., Dens, N., & De Pelsmacker, P. (2015). Is this for me? How consumers respond to personalized advertising on social network sites. Journal of Interactive Advertising, 15(2), 124-134. doi:10.1080/15252019.2015.1082450.

Dinev, T. & Hart, P. (2006). An extended privacy calculus model for e-commerce transactions. Information Systems Research, 17(1), 61-80. doi:10.1287/isre.1060. 0080.

Faul, F., Erdfelder, E., Buchner, A., & Lang, A.-G. (2009). Statistical power analyses using G*Power 3.1: Tests for correlation and regression analyses. Behavior Research Methods, 41(4), 1149–1160. doi:10.3758/BRM.41.4.1149.

French, D. P., Sutton, S., Hennings, S. J., Mitchell, J., Wareham, N. J., Griffin, S., ... & Kinmonth, A. L. (2005). The importance of affective beliefs and attitudes in the theory of planned behavior: Predicting intention to increase physical activity. Journal of Applied Social Psychology, 35(9), 1824-1848. doi:10.1111/j.15591816.2005. tb02197.x.

Hayes, A. F. (2017). Introduction to mediation, moderation, and conditional process analysis. A regression-based approach. New York: Guilford Press.

Hong, I. B. (2015). Understanding the consumer's online merchant selection process: The roles of product involvement, perceived risk, and trust expectation. International Journal of Information Management, 35(3), 322-336. doi:10.1016/j.ijinfomgt.2015. 01.003.

Kalyanaraman, S. & Sundar, S. S. (2006). The psychological appeal of personalized content in web portals: does customization affect attitudes and behavior? Journal of

Communication, 56(1), 110-132. doi:10.1111/j.1460-2466.2006.00006.x.

Kim, Y. J. & Han, J. (2014). Why smartphone advertising attracts customers: A model of web advertising, flow, and personalization. Computers in Human Behavior, 33, 256-269. doi:10.1016/j.chb.2014.01.015.

Kitchen, P. J. & Burgmann, I. (2010). Integrated marketing communication. In J. Sheth & N. Malhotra (Eds.), Wiley international encyclopedia of marketing (Vol. 4, pp. 105-128) New York: Wiley-Blackwell. doi:10.1002/9781444316568.

Kramer, A. D., Guillory, J. E., & Hancock, J. T. (2014). Experimental evidence of massive-scale emotional contagion through social networks. Proceedings of the National Academy of Sciences, 111(24), 8788-8790. doi:10.1073/pnas.1320040111. Kramer, T., Spolter-Weisfeld, S., & Thakkar, M. (2007). The effect of cultural orientation on consumer responses to personalization. Marketing Science, 26(2), 246-258. doi:10.1287/mksc.1060.0223

Kreuter, M. W., Oswald, D. L., Bull, F. C., & Clark, E. M. (2000). Are tailored health education materials always more effective than non-tailored materials? Health Education Research, 15(3), 305-315. doi:10.1093/her/15.3.305.

Lee, C. H. & Cranage, D. A. (2011). Personalisation–privacy paradox: The effects of personalisation and privacy assurance on customer responses to travel websites. Tourism Management, 32(5), 987-994. doi:10.1016/j.tourman.2010.08.011.

Lee, J. M. & Rha, J. Y. (2016). Personalization–privacy paradox and consumer conflict with the use of location-based mobile commerce. Computers in Human Behavior, 63, 453-462. doi:10.1016/j.chb.2016.05.056.

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, C. & Liu, J. (2017). A name alone is not enough: A reexamination of web-based personalization effect. Computers in Human Behavior, 72, 132-139. doi: 10.1016/j. chb.2017.02.039.

Malheiros, M., Jennett, C., Patel, S., Brostoff, S., & Sasse, M. A. (2012). Too close for comfort: A study of the effectiveness and acceptability of rich-media personalized advertising. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, 579-588. doi:10.1145/2207676.2207758.

Maslowska, E., Smit, E. G. & Van den Putte, B. (2016) It is all in the name: A study of consumers' responses to personalized communication. Journal of Interactive Advertising, 16(1), 74-85, doi:10.1080/15252019.2016.1161568.

Mitchell, A. A. & Olson, J. C. (2000). Are product attribute beliefs the only mediator of advertising effects on brand attitude? Advertising & Society Review, 1(1). doi:10. 1353/asr.2000.0010.

Moorman, C., Zaltman, G., & Deshpande, R. (1992). Relationships between providers and users of market research: The dynamics of trust within and between organizations. Journal of Marketing Research, 29(3), 314-328. doi:10.2307/ 3172742.

O'Keefe, D. J. (2003). Message properties, mediating states, and manipulation checks: Claims, evidence, and data analysis in experimental persuasive message effects research. Communication Theory, 13(3), 251-274. doi:10.1093/ct/13.3.251.

Olivero, N. & Lunt, P. (2004). Privacy versus willingness to disclose in e-commerce exchanges: The effect of risk awareness on the relative role of trust and control. Journal of Economic Psychology, 25(2), 243-262. doi:10.1016/s01674870 (02)00172-1.

from http://www.adlucent.com/blog/2016/71-of-consumers-prefer-personalized-ads/. Percy, L., & Rossiter, J. R. (1992). A model of brand awareness and brand attitude

advertising strategies. Psychology & Marketing, 9(4), 263-274. doi:10.1002/mar. 4220090402.

Petty, R. E. & Cacioppo, J. T. (1986). The elaboration likelihood model of persuasion. In Communication and persuasion (pp. 1-24). New York: Springer. Retrieved from http://www.psy.ohiostate.edu/petty/documents/1986ADVANCESsPettyCacioppo.pdf Rimer, B. K., & Kreuter, M. W. (2006). Advancing tailored health communication: A persuasion and message effects perspective. Journal of Communication, 56(1), 184-201. doi:10.1111/j.1460-2466.2006.00289.x.

Roderick, L. (2017, June 29). ‘Arrogance’ around brand purpose making consumers distrust ads. Retrieved from https://www.marketingweek.com/2017/ 06/29/arrogance-brand-purpose-distrust-ads/.

Simonson, I. (2005). Determinants of customers’ responses to customized offers: Conceptual framework and research propositions. Journal of Marketing, 69(1), 32-45. doi:10. 2139/ssrn.405060.

Smit, E. G., Van Noort, G., & Voorveld, H. A. M. (2014). Understanding online behavioural advertising: User knowledge, privacy concerns and online coping behaviour in Europe. Computers in Human Behavior, 32, 15-22. doi:10.1016/j.chb.2013.11.008.

Song, J. H., Kim, H. Y., Kim, S., Lee, S. W., & Lee, J. H. (2016). Effects of personalized e-mail messages on privacy risk: Moderating roles of control and intimacy. Marketing Letters, 27(1), 89-101. doi:10.1007/s11002-014-9315-0. Tu, F. (2016). WeChat and civil society in China. Communication and the Public,1(3),

Tversky, A., & Kahneman, D. (1981). The framing of decisions and the psychology of choice. Science, 211(4481), 453-458. doi:10.1007/978-1-4613-2391-4_2.

Walrave, M., Poels, K., Antheunis, M. L., Van den Broeck, E., & Van Noort, G. (2016). Like or dislike? Adolescents’ responses to personalized social network site advertising. Journal of Marketing Communications, 1-18. doi:10.1080/13527266. 2016.1182938.

Wu, G. & Wu, G. (2006). Conceptualizing and measuring the perceived interactivity of websites. Journal of Current Issues & Research in Advertising, 28(1), 87-104. doi:10.1080/10641734.2006.10505193.

Yang, B., Kim, Y., & Yoo, C. (2013). The integrated mobile advertising model: The effects of technology-and emotion-based evaluations. Journal of Business Research, 66(9), 1345-1352. doi:10.1016/j.jbusres.2012.02.035.

Yu, J. H. & Cude, B. (2009). ‘Hello, Mrs. Sarah Jones! We recommend this product!’ Consumers' perceptions about personalized advertising: Comparisons across advertisements delivered via three different types of media. International Journal of Consumer Studies, 33(4), 503-514. doi:10.1111/j.1470-6431.2009.00784.x.

Zaichkowsky, J. L. (1994). The personal involvement inventory: Reduction, revision, and application to advertising. Journal of Advertising, 23(4), 59-70. doi:10.1080/ 00913367.1943.10673459.

Appendix A Pilot study

A pilot study with 38 participants, mainly university students, was conducted. To determine the different levels of personalisation, different types of personal data were evaluated by the item “To what extent do you mind if these information categories will be used for advertising purposes?” and measured on a 7-point Likert scale anchored by 1 (extremely inappropriate) and 7 (extremely appropriate). The results are presented in Table A1. Due to these means and standard deviations as well as due to practical reasons, gender (M = 5.77, SD = 1.33) was chosen as the low, location (M = 4.29, SD = 1.85) as the moderate and credit card number (M = 1.76, SD = 1.89) as the high personalisation condition.

The second section aimed to determine the levels of involvement attached to different products. The items tested were based on the findings of the Rossiter Percy grid, which categorises product types into high and low involvement products (Wu & Wu, 2006; Percy & Rossiter, 1992). Thus, two high and two low involvement products were tested in the pilot study. The participants were exposed to pictures of the product. As being very close to the final stimulus material, this simultaneously depicted a pre-test of the stimulus material (see Figure B1&B2). The consumers’ level of involvement was assessed on a 7-point Likert scale anchored by 1 (strongly disagree) and 7 (strongly agree). The seven items used (Hong, 2015) were based on Zaichkowsky’s (1994) revised personal involvement inventory. One example was: “This product is important to me”. For the remaining items see Table C1. The product evoking the highest level of involvement was travels (M = 5.00, SD = 1.70) and the one evoking the lowest level of involvement was painkillers (M = 2.96, SD = 0.99), see Table A2. Accordingly, travels and painkillers were chosen as products for manipulating the involvement conditions.

Table A1

Consumers’ approval of different types of personal data for advertising

Type of personal data M SD

age gender education occupation marital status name e-mail address telephone number home address shopping behaviour personal interests religious belief political views location income

credit card number

5.60 5.77 4.47 4.35 3.65 2.88 3.26 1.85 2.00 5.00 4.94 2.88 2.68 4.29 2.71 1.76 1.36 1.33 1.31 1.43 1.41 1.82 1.90 1.08 1.26 1.72 1.81 1.36 1.36 1.85 1.93 1.89

Note. All outcomes were rated on a 7-point Likert scale with 1 as lowest and 7 as highest value.

Table A2

Consumers’ product involvement

Product type M SD travels insurance candy painkillers 5.00 3.68 3.71 2.96 1.70 1.18 1.34 0.99