A quantitative study on self-disclosure in online advertising

(1)

A Quantitative Study on Self-Disclosure in Online Advertising

Author : Daan van der Hofstede University of Amsterdam

Faculty of Business and Economics Amsterdam Business School Executive Programme in Management Studies

Track : Digital Business

Thesis Supervisor : Dr. Hüseyin Güngör Date of Submission : 27.06.2018

(2)

2

Statement of Originality

This document is written by Student Daan van der Hofstede who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

3

ABSTRACT

Although the amount of personal information which is provided to firms in online advertising

is growing rapidly, the reasons why people decide to self-disclose in these kind of online

environments have not been adequately investigated. This study builds a model in which design similarity and pre-filling of personal information in webforms are being related to self-disclosure. In addition, this study has developed a framework to investigate the effects of select drivers of consumers deciding to self-disclose in order to get access to exclusive discount vouchers for online shopping purposes. Our findings from two different studies based on primary and real secondary data reveal that the probability of self-disclosure is the highest in case 1) the design of the advertising platform looks similar to the website that was visited prior to the moment that the consumer got redirected to the external advertising platform and 2) the personal information of the consumer was already pre-filled in the webform that needs to be submitted to self-disclose. This research shows that, except for trust, personal relevance and brand awareness, the identified drivers namely, privacy concerns, perceived benefits and the customer effort score significantly influence the consumers' willingness of self-disclosure.

Keywords: Self-disclosure; Privacy Concerns; Design Similarity; Perceived Benefits,

(4)

4

1. INTRODUCTION

The amount of personal data shared in an online environment is growing by the day. Data creates value for the global economy, driving innovation, productivity, efficiency, and growth (Stanford Law, 2012). Personal data refers to information which can be directly linked to an individual e.g. gender, name, e-mail address, IP address, order ID’s and more. In order to make the internet safer, to guarantee privacy and to better protect personal data, a new law has been

adopted at European level that has become effective on the 25th_{of May 2018: General Data}

Protection Regulation (GDPR) (DLAPIPER, 2017; Voss, 2017). Consequently, firms are not allowed to store personal data and send any form of personalized communication without explicit consent from the consumer. This form of consent is given at the moment the consumer shares his personal information by completing a registration on the internet. Furthermore, personal data and pseudo-anonymous data may only be used for specified, explicit and legitimate purposes and not further processed in ways that are not compatible with these purposes (Takens, 2017). The GDPR applies to all firms that collect or process personal data of EU citizens. (DLAPIPER, 2017). Experts in the field of online marketing state that new privacy legislation coming with the GDPR will cause significant damage on the online marketing industry (O’Reilly, 2015). Taking this into account, a solid approach to simultaneously meet legal requirements while getting explicit consent from the consumer has become much more important to data-driven businesses and in particular firms active in online marketing. Besides the fact that explicit consent is having a positive impact on consumers' responses towards interactive marketing activities (Tsang, Ho, & Liang, 2004), explicit consent is now also a legal requirement and therefore, crucial for companies that use direct communication for marketing purposes (Krafft, Arden, & Verhoef, 2017).

(6)

6 Nowadays, potential customers can be targeted and contacted on an individual level.

Personalized and relevant communication often shows higher conversion ratios1_{. “A conversion}

occurs when a visitor to a website completes a desired goal, such as filling out a form or making a purchase” (Wordstream, 2018, p.1). A higher conversion ratio directly benefits the firm and its online marketing effort (Schumann, von Wangenheim, & Groene, 2014). However, prior research showed that consumers often perceive personalized messages based on personal preferences as an intrusion of their privacy (Awad & Krishnan, 2006; Krafft et al., 2017). Jolley, Lee, Mizerski, & Sadeque (2013) showed that explicit consent from the consumer increases the effectiveness of personalized communication sent by firms and even has a positive effect on the retention of existing customers. A higher degree of personalization might also cause increased privacy concerns (Tucker, 2014). Walrave & Heirman (2012) found that privacy concerns and perceived benefits do explain a considerable portion of variance related to the willingness of consumers’ to disclose personal information in an online environment. Prior research has shown that decisions related to self-disclosure of personal information also depend on a cost-benefit analysis of the consumer. This is underlined by the social exchange theory of Homan (1961) and the personalization privacy paradox. The personalization privacy paradox explains the inconsistency between the expressed concerns and actual consumer behavior. The social exchange theory states that people only participate in an exchange at the moment they estimate in advance that the outcome will be positive.

Until now, research related to the social exchange theory and the personalization privacy paradox related to the willingness of sharing personal information has been applied to multiple work fields. So far, researchers have primarily focused on responses to personalized marketing, effects of privacy concerns and on opt-in decisions. To the best of our knowledge there is still a lack of academic research that develops and tests a conceptual model with a large number of

(7)

7 theory-based constructs related to self-disclosure decisions in online advertising. There is yet no study with a specific focus on how self-disclosure decisions are related to website design and prefilling personal details during the registration process. This study is aiming to examine how sharing personal information with an online advertising platform is influenced by design similarity and pre-filling personal information. Another essential point is how self-disclosure decisions are influenced by theory-based constructs such as trust, customer effort score, personal relevance, brand awareness, perceived benefits and privacy concerns. This leads us to the following research question:

How do design similarity and pre-filling personal information influence the consumers’ willingness of self-disclosure and how can this be explained?

In the following chapter we consider and integrate insights from this section's discussion of theoretical approaches and relevant literature on self-disclosure to substantiate our conceptual framework in chapter three. In this research secondary data from an online advertising platform was used as an empirical basis (study A) for our research based on primary data (study B). In chapter four we provide background information about the online advertising platform together with the experimental manipulations to conclude with the results of study A. These results are foundational to our main research which is explained in chapter five. In study B we have reproduced the different scenarios coming from study A to see if the results from both studies would have the same outcome and to identify what factors are influencing self-disclosure decisions. Results of study B are presented in chapter six followed by the discussion, limitations and suggestions for further research in chapter seven.

(8)

8

2. LITERATURE REVIEW

So far, researchers have primarily focused on responses to personalized marketing and effects of privacy concerns. Limited research has been conducted on how self-disclosure decisions in online advertising are influenced by website design and pre-filling personal information during the registration process. Current knowledge about self-disclosure and privacy concerns was closely examined as prior research showed that these two factors are highly correlated. To give a good idea of the current knowledge in the area of self-disclosure and privacy concerns, we have included the overview showed in table 1. Next, a definition of the concept self-disclosure and the principles of privacy calculus, social exchange theory and the personalization privacy paradox will be introduced followed by the link between the theory based constructs and self-disclosure. Contained in these sections is a set of hypotheses that will be explored as part of this study. In the next chapter a visual representation of this study’s conceptual framework is presented.

(9)

9

Table 1: Current knowledge within the field of self-disclosure and privacy concerns

2.1 Self-Disclosure

The process of making yourself known to others is known as self-disclosure (Jourard and Lasakow, 1958). “Disclosure between an individual and an organization can serve authentication purposes – for instance, to establish identity, allow authentication of a claim to identity and to enable an organization to recognize you in the future in order to personalize its offerings to you. Organizations might also ask for personal information for marketing purposes – for instance, when registering to access a website or joining an online community” (Joinson & Paine, 2007, p. 236). In this research we will focus on the phenomenon of providing personal information (salutation, name and e-mail address) to an online form via webforms.

2.1.1 Cost-Benefit Trade-Off

There is a rich body of literature that is linking self-disclosure decisions on the internet to privacy concerns (Krafft et al., 2017; Norberg, Horne, & Horne, 2007; Taddicken, 2014; Tucker, 2014). Different theories have been established mentioning that cost-benefit / risk-benefit trade-offs are the basic principle when it comes to sharing personal information online. From a consumer perspective, this basic psychological approach seems to be foundational with regard to information exchange. To provide a deeper understanding of the most common theories related to cost-benefit trade-offs in the field of information exchange, a brief explanation of the different concepts is included.

2.1.2 Privacy Calculus

“A subset of empirical studies introduces the concept of privacy calculus by assuming that a consequentialist tradeoff of costs and benefits is salient in determining an individual’s behavioral reactions ” (Smith, Dinev, & Xu, 2011, p.1001). This concept which is referred to

(10)

10 as the privacy calculus, states that if consumers are requested to provide personal information to a firm, a risk-benefit analysis will be performed to assess the return for the information that is being asked for. As a result, this risk-benefit trade-off will eventually lead to the decision whether or not to provide personal details. The privacy calculus theory assumes that a consumer only provides personal information when the benefits outweigh the risks (Chellappa & Sin, 2005; Culnan, 1993; Dinev & Hart, 2006; Hoadley, Xu, Lee, & Rosson, 2010; Hui, Tan, & Goh, 2006; Milne & Gordon, 1993; Milne & Rohm, 2000).

2.1.3 The Social Exchange Theory

The cost-benefit trade-off that has been mentioned in the previous section is also being referred to within the social exchange theory by Homans’ (1961). “This theory explains that humans only decide to engage in an exchange situation if they expect the net outcome to be positive. Social exchange theory serves to explain the basics of human interaction and has been frequently applied in the context of information exchange (Culnan & Armstrong, 1999; Schumann et al., 2014)” (Krafft et al., 2017, p.40).

2.1.4 Personalization Privacy Paradox

Firms are increasingly interested in what factors are having significant effects on the consumers’ decision to (not) share their personal information. In case a consumer is willing to interact with a firm via direct online communication, personal information has to be shared with explicit consent. Hence, accepting general terms and conditions including privacy statements during the registration process is mandatory. Consequently it is all or nothing; either the consumer accepts and completes the registration or the process will be aborted. Self-disclosure in an online environment causes firms to have the possibility to show advertisements to its customers, sending them newsletters about new products and to personalize the customer

(11)

11 journey on an individual level (Lemon & Verhoef, 2016a). From a firm perspective, without having the right to process personal data these kind of marketing tools cannot be used.

Interestingly, consumers seem to behave inconsistent to what they express about privacy related to their willingness to disclose personal information on the internet (Acquisti & Grossklags, 2004; Norberg et al., 2007; Taddicken, 2014). Acquisity and Grossklags (2004) found that consumers still provide their personal details despite the fact that they have mentioned to be concerned about their privacy. Remarkably, they often do not put a lot of effort in securing their personal data. This contradiction is known as the personalization privacy paradox. One of the many underlying explanations for this inconsistency is that consumers are willing to trade their personal information for advantages (e.g. rewards, monetary incentives or exposure (Acquisti & Grossklags, 2004)). To illustrate, Xu, Luo, Carroll, & Rosson (2011) found that consumers who express that they are concerned about their online privacy, often still disclose their personal information in return for added value such as personalized ads that will be delivered based on their context.

2.2 Privacy Concerns

“If this is the age of information, then privacy is the issue of our times.” (Acquisti, Brandimarte, & Loewenstein, 2015, p.509). Privacy can be defined as the right to be left alone (Dinev, Xu, Smith, & Hart, 2013). Following the definition in the field of online advertising privacy concerns refer to the consumers awareness of losing privacy as a consequence of disclosing personal information while engaging in consent-based data exchange with a company (Krafft et al., 2017; Tucker, 2014). Many researchers found that privacy concerns are closely related to the consumers’ reluctance of sharing personal data (Krafft et al., 2017; Norberg et al., 2007; Taddicken, 2014). Over the last decades the use of personal data has created tension between firms economic interests and consumers’ privacy concerns. In times of big data and machine learning companies are gathering as much consumer information as possible in order to sustain

(12)

12 their competitive advantage (Biesdorf, Court, & Willmott, 2013; Culnan & Bies, 2003). On the other hand, concerns about privacy violations became more important as consumers cannot control the secondary use of their personal data (Norberg et al., 2007).

2.2.1 Factors That Affect Privacy Concerns

Loss of control over personal data management is being considered as one of the major reasons for consumers to have a negative attitude towards sharing personal information (Laufer & Wolfe, 1977). As a follow-up to the cost-benefit trade-off that has already been mentioned in the section above, the loss of privacy can be considered as a cost factor. This is because personalized communication as a result of self-disclosure can be seen as potentially intrusive (Tucker, 2014; Van Doorn & Hoekstra, 2013). Perceived control over personal data management has been found to be positively related to the probability that a consumer is exchanging personal data with a company on the internet (Krafft et al., 2017). Malhotra, Kim, & Agarwal (2004, p.350) stated that “online consumers consider it most important to 1) be aware of and 2) have direct control over personal information stored in marketers' databases”.

By self-disclosing personal information to an online advertising platform consumers agree that their behavior on the internet will be followed. Online tracking systems do follow clickstreams and browsing behavior to adjust the customer journey on an individual level. Research in the field of online advertising showed that a higher level of personalization leads to higher conversion ratios (Schumann et al., 2014). In case communication becomes too personal this has a negative effect on privacy concerns (Chellappa & Sin, 2005; Culnan & Armstrong, 1999; Krafft et al., 2017).

Martin, Borah, & Palmatier (2017) reported three studies showing that transparency and control in firms data management practices can suppress the negative effects of customer data vulnerability. “However, such efforts also increase perceptions of susceptibility to harm due to unwanted uses of their personal data, such as those that can result from data breaches or identity

(13)

13

theft” (Martin et al., 2017, p.36). Just take the recent examples of Facebook2_{(Kelly, 2018)}

where personal data was misused to influence elections or the recent data breach at

MyFitnessPal3_{(Agencies, 2018)} _{which caused the leakage of data coming from over 150}

million profiles. Factors like transparency and control, separately and interactively, mitigate the damaging effects of all types of customer data vulnerability (Baumeister, Zhang, & Vohs, 2004). This includes the positive effect on privacy concerns and the negative effect on trust. Also, research done by Xie, Teo, & Wan (2006) has mainly shown that specific factors, such as trust, reduce the negative effect of privacy concerns on decisions to share information. Interestingly, factors like company reputation, consumer-sided trust, and data protection seals can increase the level of trust and attenuate the negative impact of privacy concerns (Xie, Teo, & Wan, 2006).

Related to the cost-benefit trade-off consumers are usually reluctant to provide personal information or tend to provide false information online because of their concern about privacy violation (Xie et al., 2006). Privacy concerns could eventually lead to advertising avoidance (Baek & Morimoto, 2012). Prior research done by Krafft et al. (2017) has shown that the combined (interaction) effects of perceived benefits with privacy concerns are in line with insights from related streams of literature (Awad & Krishnan, 2006; Edwards, Li, & Lee, 2002; Van Doorn & Hoekstra, 2013). Krafft et al. (2017) also shows that if consumers are concerned about their privacy, the negative effects can at least be compensated for by designing interactive marketing content that fits their needs.

In the past, research has already been conducted into the consequences of privacy concerns on the use of online interactive media (Demoulin & Zidda, 2009; Zhao, Lu, & Gupta,

2_{http://money.cnn.com/2018/04/04/technology/facebook-cambridge-analytica-data-87-million/index.html} 3

(14)

14 2012). Rewards, in the form of a monetary voucher, were found to have a positive impact on online consumers' decision to provide accurate personal information for personally identifiable data but not for demographic data (Xie et al., 2006). Apparently, consumers are increasingly aware of the value of their personal information in the context of interactive media (Tsai, Egelman, Cranor, & Acquisti, 2011). Based on these arguments, we assume:

H1: The higher the perceived privacy concerns, the lower the probability that consumers

will self-disclose.

2.3 Trust

“Using the internet involves a leap of faith. We type in our credit card numbers and other personal information in order to make purchases over the internet and trust that this information will not be used in unauthorized or fraudulent ways” (Bargh & McKenna, 2004, p.585). The extent to which consumers believe that the internet is a safe environment to process and exchange personal information is strongly related to privacy concerns (Schlosser, White, &

Lloyd, 2006). Thus, it is important to understand how firms can communicate their

trustworthiness to consumers. Marketing managers often find difficulties with establishing consumers’ trust in a broad range of topics, however implementing this trust factor in computer-mediated environments like the internet may be may be challenging (Naquin & Paulson, 2003). A common approach is to add statements to websites that make clear that personal data are used and warranted discreetly (e.g., privacy and security statements). These statements are often placed at the far bottom of a website and from a consumer perspective difficult to understand. Some research has shown that such statements help instill consumer confidence in advertising sites (Palmer, Bailey, & Faraj, 2000), others suggest that they are not necessarily the most important predictor of online trust (Montoya-Weiss, Voss, & Grewal, 2003; Shankar, Urban, & Sultan, 2002). “Findings from a recent large-scale study suggest that despite consumers’ claims that privacy policies are important for establishing credibility, consumers refer instead to

(15)

15 “surface” elements, such as website design (Fogg et al., 2003).” (Schlosser et al., 2006, p.133). Factors such as reputation and data protection statements provide trust and weaken the negative consequences of privacy concerns (Xie et al., 2006).

Because these findings support the existence of a positive impact of trust on the willingness of self-disclosure, we conclude:

H2: The higher the level of trust, the higher the probability that consumers will self-disclose.

2.4 Customer Effort Score

The perceived cost of the registration process in terms of time and effort is what we call the ‘customer effort score’. A long, user unfriendly and complicated registration process is likely to have a negative effect on the probability that the consumer will finish the registration. Krafft et al. (2017) found that complicated phrasing and long-winded terms and conditions during registration is positively related to consumers aborting the registration process.

“Optimized web forms lead to faster completion times, less form submission trials, fewer eye fixations and higher user satisfaction in comparison to the original forms” (Seckler, Heinz, Bargas-Avila, Opwis, & Tuch, 2013, p.1893). However, we have to be aware that research done on customer effort scores related to web forms is both limited and rather old. The perceived customer effort score is often dependent of the amount of information which is required to register. Because of this it is assumed that the amount and the complexity of the required information are negatively related to the probability to successfully complete the registration process (Dickinger, Haghirian, Murphy, & Scharl, 2004; Krishnamurthy, 2001).

H3: The higher the perceived level of effort, the lower the probability that consumers

(16)

16

2.5 Personal Relevance

“The need for personally relevant information can be identified as a main factor that drives consumers to interact with a company. Such information represents substantial value and, thus, positively affects consumers' willingness to provide personal information.” (Krafft, Arden, & Verhoef, 2017, p.42). From a consumer perspective; a higher degree of personalization in marketing activities is having a positive effect on the perceived level of relevance (Ansari & Mela, 2003). Prior research offers some evidence that a consumers' reaction to a customized message depends on both the degree of customization and the potential benefits of the offer (Van Doorn & Hoekstra, 2013). Higher degrees of personalization in terms of communication towards a consumer at the same time has a positive effect on the intention to purchase (Van Doorn & Hoekstra, 2013). “Consumers might see personalized ad content on such sites as more appealing and more connected to their interests, but they also may see it as ‘not only creepy, but off-putting’ if they feel that the firm has violated their privacy. These privacy concerns may lead to ‘reactance’ which leads consumers to resist the ad’s appeal” (Tucker, 2014, p.1).

Feeling special, receiving personalized communication, being different, having the option to determine for yourself how to interact with a company. Personalization of the customer journey, it’s becoming more popular by the day. Prior research has shown that the consumers’ willingness to interact with a corporation is higher if they do expect to receive information and/or offers on an individual level (Berry, 1995; Gwinner, Gremler, & Bitner, 1998). Changing online touchpoints to interact on a more personal level is on the shortlist of many companies (Lemon & Verhoef, 2016). The allocation of IT resources to adjust and increase the interactivity of the customer journey is in this case closely related to increasing customer engagement and brand awareness (Lemon & Verhoef, 2016). On the other hand, a study done by Baek & Morimoto (2012) showed that perceived personalization of a customer journey can also have a negative effect on advertising-skepticism and advertising-avoidance.

(17)

17 Although it was tested in a different marketing context (location based marketing) research done by Zhao et al. (2012) showed that the personalization of the customer journey has a positive effect on the probability that a consumer is willing to disclose personal data in its interaction with a company. Because relevance can also be closely related to trust (Baek & Morimoto, 2012; Martin et al., 2017) it might be that the probability that the consumers’ willingness of self-disclosure is influenced by personal relevance. We hypothesize:

H4: The higher the perceived personal relevance of discount vouchers, the higher the

probability that consumers will self-disclose.

2.6 Brand Awareness

Brand awareness means the ability of a consumer to recognize and recall a brand in different situations (Aaker, 1996). Brand awareness consists of brand recall and brand recognition (Chi, Yeh, & Yang, 2009). Brand recall relates to situations in which a consumer can directly appoint a certain brand. Brand recognition means a consumer has the ability to identify a brand when there is a brand cue (Chi, Yeh, & Yang, 2009). Advertising creates and enhances brand awareness by exposing brands to customers (Aaker & Keller, 1990; Batra, Lehmann, Burke, & Pae, 1995; Keller, 1993; Rossiter & Percy, 1987; Yoo, Donthu, & Lee, 2000). Brand awareness also acts as a critical factor in the consumers’ online decision making process (Chi, Yeh, & Yang, 2009). As we have discussed above, self-disclosure decisions in the online advertising industry are closely related to privacy concerns and trust. Little research has been done on how self-disclosure decisions are influenced by brand awareness. For instance, the influence of brand awareness on decision making is explored using only lab experiments at the individual

consumer level(Huang & Sarigöllü, 2014; Macdonald & Sharp, 2000). A study done by Aaker

& Keller (1990) showed that the higher level of brand awareness the higher the brand trust. Recently, the rising of consumer consciousness has ensured that consumers are more likely to interact with online platforms they are more familiar with (Chi, Yeh, & Yang, 2009). For that

(18)

18 reason consumers might have a more positive attitude towards advertising platforms if they do recognize the brand which is facilitating the service. Thus, we hypothesize:

H5: The higher the level of brand awareness, the higher the probability that consumers

will self-disclose.

2.7 Perceived Benefits

“Perceived benefits has been characterized as a primary motivation for entering into marketing relationships and the essential outcome of marketing activity (Babin et al., 1994; Holbrook, 1994)” (Forsythe, Liu, Shannon, & Gardner, 2006, p.57). Following the idea of the privacy calculus we have included perceived benefits as one of our theory-based constructs. Consumers are known to be motivated to maximize benefits and minimize risks (Forsythe, Liu, Shannon, & Gardner, 2006), both the perceived benefits and risks are expected to play important roles in explaining consumers’ willingness of self-disclosure. “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).” (Ahrens, 2018, p.7). Prior research in the field of location aware marketing did relate perceived benefits to self-disclosure decisions. One of the findings was that consumers are likely to agree to give up a degree of privacy in return for potential benefits related to information disclosure (Xu, Luo, Carroll, & Rosson, 2011). However, literature on self-disclosure decisions related to perceived benefits in the field of online advertising is both limited and rather old. Because the perceived benefits are expected to be always positively correlated (Xu, Teo, Tan, & Agarwal, 2009) we hypothesize:

H6: The higher the level of perceived benefits, the higher the probability that consumers will self-disclose.

(19)

19

2.8 Design Similarity & Pre-filling Personal Information

On a daily basis our decisions to disclose personal information in online environments are influenced by a number of factors, some of which we may not be even aware of. The graphical design of a customer journey is one of those factors. “Psychology researchers call it fluency while web developers call it usability, but they’re both basically talking about the same thing — how well something is designed can directly impact how much people use it. And not only the degree to which they use it, but also the amount of self-disclosure a person makes while using it.” (Grohol, 2018, p.1). This implies that the graphical design of the customer journey is one of the drivers that can affect the decision to self-disclose. A study done by Alter, Oppenheimer, & Epley (2013) demonstrated that the probability of self-disclosure is higher under high-fluency conditions: The easier it is for consumers to process information during the registration process, the more likely it is that they decide to self-disclose. Alter et al. (2013) also found that consumers process information more deeply, but not necessarily more accurately, when disfluency is experienced. In addition, disfluency can trigger more analytical thinking.

In our research we have included two experimental variables called ‘design similarity’ and ‘pre-filling personal information’. Design similarity relates to the graphical design of the customer journey where pre-filling personal information refers to webforms which are already filled with the personal details of the consumer at the moment the webform is presented. Both variables will be described in more detail in chapter four of this study as they relate to our manipulations in our experiment. Following the information provided in the first paragraph, we expect both manipulations to have an effect on the level of perceived fluency; if the graphical design of the customer journey changes from one click to another, it might be that this increases the level of perceived disfluency. Furthermore, it might also be that more analytical thinking is triggered which increases the level of perceived privacy concerns; especially in case the website

(20)

20 is asking for personal information. Pre-filling personal information during the self-disclosure process is often used to increase the level of usability which in that sense is proved to have a positive effect on fluency (Johnson, Bellman, & Lohse, 2002; Lemon & Verhoef, 2016) and self-disclosure (Ahrens, 2018; Krafft, Arden, & Verhoef, 2017). Following those arguments and findings, we postulate:

H1a & H7: Compared to a design which is not similar, the willingness of self-disclosure

is higher in case the design of the customer journey is similar.

H8: A similar design of the customer journey with having the personal details from the

consumer pre-filled, will show the highest probability of self-disclosure.

H2a & H9: A design of the customer journey which is not similar without having the

personal details from the consumer pre-filled, will show the lowest probability of self-disclosure.

H3a: Switching the design from not similar to similar will have a positive effect on the

consumers’ willingness of self-disclosure.

3 CONCEPTUAL FRAMEWORK STUDY A + B

In this research we focus on different theory-based constructs and their explanatory power on our dependent variable; the consumers’ willingness of self-disclosure.

As explained in the literature review the research by Krafft et al. (2017) distinguished between more economic and psychological benefits and costs. In this research perceived benefits relate to the option to choose an exclusive shopping-voucher while psychological costs involve the actions or data input the consumer need to undertake (the customer effort score) before self-disclosure. Although the chosen set of variables do not cover the full picture of what parameters could influence the consumers’ willigness of self-disclosure we have included most

(21)

21 important factors coming out of prior research. To the best of our knowledge these variables have never been tested together with a specific focus on online advertising.

We base our inclusion of cost-related factors mainly on the literature mentioned in the upper section. It can be assumed that consumers drop out during the self-disclosure process if too much personal information is requested (Dinev & Hart, 2006; Johnson, Bellman, & Lohse, 2002; Martin et al., 2017). Furthermore, the feeling that personal data may not be dealt with properly is also expected to have a negative impact on self-disclosure (Martin et al., 2017). The actual design of this study can be captured as following (see figure 1):

4. STUDY A: EMPIRICAL BASIS

In this study real secondary data from an online advertising platform was used as an empirical basis for further research. This data is based on an experiment that was conducted to see if the design of the customer journey and the presence of pre-filling personal information in webforms during the registration process, would influence the probability of self-disclosure.

This advertising platform facilitates a closed network of online shops covering all consumer market segments. The concept of this service for firms active in the e-commerce

(22)

22 segment is to reward consumers for their purchase after they have completed the checkout procedure in an online shop. The process after a consumer has bought something in one of the participating online shops can be captured in the following steps:

1. A consumer buys online in one of the shops connected to the advertising platform. After the order has been paid for and the customer journey has been completed, the consumer will land on the check-out page where a summary of the order details is being provided. For instance, normally this page shows an order-ID and confirms that the purchase has been completed successfully. The order will now be processed by the online shop. 2. On this check-out page the consumer is exposed to a reward-banner as a reward for

purchasing. The reward-banner displays a present box together with a short teaser: “A

small gift is waiting for you! As a thank you for shopping with us, you can choose an

exclusive voucher from our network of partner shops.”

3. Clicking on the reward-banner will lead the consumer to an external website where a wide variety of exclusive discount-vouchers of the other participating shops will be displayed.

4. The consumer can only select one discount-voucher. Before the consumer gets access to the discount-voucher personal data, in this case, salutation, first name, last name and e-mail address) have to be disclosed. Without self-disclosure it is not possible to request the voucher.

5. To finalize the request the consumer needs to give explicit consent to the privacy statement and general terms and conditions.

We have included the following visual for a better understanding of the customer journey of the advertising platform:

(23)

23

Figure 2. Visualization of the customer journey

4.1 Research Methodology

We conducted an empirical study to gain more knowledge about what factors influence the level of self-disclosure. The first section outlines the sample and the main characteristics of the data. In the second part we go deeper into the dependent variable and conclude with the analysis methodology.

Prior to the experiment the design of the customer journey, which starts after the consumer clicked on the reward-banner, was different in comparison to the design of the online shop. Therefore, it was obvious to the consumer that the service was offered by an external advertising platform. The experiment that has been conducted did investigate if the probability that a consumer will decide to self-disclose on the advertising platform could be increased by adjusting the graphical design of the customer journey to the same look and feel of the online shop where the consumer initially did the purchase. In addition, it was also tested if pre-filling the personal information of the consumer in the webform that is used to request a voucher would have a positive impact on the probability of self-disclosure. Consequently, the experiment consisted of four different scenarios:

1. Design similar, not-prefilled

(24)

24 3. Design not similar, not pre-filled

4. Design not similar, pre-filled

A visualization of the four different scenarios is presented in figure 3.

Figure 3. Visualization of the four different scenarios

In the conceptual model of this study we assumed that if the design of the customer journey of the external advertising platform would be adjusted to the look and feel of the shop where the consumer initially did the purchase, this would have a positive effect on the consumers’ willingness of self-disclosure. Furthermore, it was also assumed that if the personal details of the consumer were already pre-filled in the webform to request the discount voucher, this would also have a positive effect on self-disclosure.

(25)

25

4.2 Manipulation

The one and only manipulation in this experiment that was used to test hypothesis H3a, is about switching the design from not similar in the control or ‘before’ group to similar in the experimental or ‘after’ group. This to see if this would affect the probability of self-disclosure. However, this could only be done for scenarios three and four as the design for scenario one and two was already similar. Therefore, the experimental conditions between the control and experimental group remained the same for scenario one and two but did change for scenario three and four.

To clarify, the conditions related to the pre-fill of personal details did not change between the control and experimental groups.

4.3 Data Characteristics and Sample

Real data from 141 different online shops that used the service of the advertising platform during the time that the experiment was conducted was taken into our analysis. All four scenarios were covered with a minimum of 25 online shops per scenario. The control or ‘before’ group consisted of data collected during a timeframe of 3 months from January until March 2018. The experiment lasted one month, the data used for the experimental group was collected during April 2018. To increase the reliability of this experiment 29 online shops were removed from the dataset as those online shops did generate less than a thousand sales during April 2018. The amount of sales per shop is an important factor because the consumer will only be exposed to the reward-banner on the checkout page after the purchase has been completed successfully. The sample used for the analysis consisted of data generated by 112 different online shops (N=112). In total the control groups consisted of 3.678.267 consumers that were exposed to the reward-banner on the different check-out pages. For the experimental groups the total amount of reward-banner impressions was 1.129.781. Ultimately, only the consumers who clicked on the reward-banner and got redirected to the advertising platform were taken into the analysis.

(26)

26

4.4 Measurement of the Dependent Variable

The dependent variable ‘self-disclosure’ is binary as the decision to provide your personal information to the advertising platform followed by clicking on the submit button can be answered with ‘yes = self-disclosure’ or ‘no = no self-disclosure’. Because data was available on shop level, no individual cases could be taken into the analysis. Hence, we used the self-disclosure averages per shop to run our analysis. For this reason each shop in both the control and experimental group showed a self-disclosure value between 0 (no self-disclosure) and 1 (self-disclosure).

4.5 Results Testing hypothesis 1a

To test H1a a one-sample t-test was conducted. We hypothesized that a customer journey with a similar design would show a higher probability of self-disclosure in comparison to a customer journey which is not similar. We compared the probability of self-disclosure between the control groups of scenario one and two (similar design) versus scenario three and four (design not similar). Below the descriptive statistics related to the self-disclosure averages and the results of the one-sample t-test are provided.

(27)

27 We found a statistically significant difference in the scores for self-disclosure between the design similar group (M=0,2552, SD=0,0747) and the design not similar group (M=0,0983, SD=0,0385) conditions; t (51)=15,14, p = .000. These results show that the probability of disclosure in the group with a similar design is significantly higher than the probability of self-disclosure in the group without. Therefore, H1a is supported.

Testing hypothesis 2a

We hypothesized that a customer journey with a design which is not similar and without having the personal details from the consumer pre-filled, would show the lowest probability of self-disclosure. We compared the probability of self-disclosure between the control groups of all different scenarios. Below the descriptive statistics related to the self-disclosure averages are provided:

In line with our expectations scenario 3; design not similar & not pre-filled shows the lowest probability of self-disclosure (M=0,0711, SD=0,0254) compared to scenario 1; design similar & not pre-filled (M=0,2004, SD=0,0236), scenario 2; design similar & pre-filled (M=0,3143, SD=0,0652) and scenario 4; design not similar & pre-filled (M=0,1339, SD=0,0178). Thus, H2a is supported. A design which is not similar to the design of the online shop without having personal details from the consumer pre-filled shows the lowest probability of self-disclosure.

(28)

28

Testing hypothesis 3a

To test H3a we conducted a set of one-sample t-tests to see if switching the design of the customer journey from design not similar to similar would have a statistically significant effect on the probability of self-disclosure. As described in § 4.1 & § 4.2 we used both the control and experimental groups to test the effect of our manipulation. For scenario three and four the design of the customer journey was not similar in the control groups. In the experimental groups we applied the manipulation by switching the design from not similar to similar to test if there are significant differences among the control and experimental groups of scenarios three and four. In addition, as the design of the customer journey in scenario one and two was already similar in the control groups, the conditions did not change for the experimental groups. Therefore, we expected not to find any statistically significant differences when comparing the results of the control and experimental groups for scenario one and two.

Scenario 1 and 2: Design similar, not prefilled & pre-filled

A one-sample t-test was conducted to compare the probability of self-disclosure between the control and experimental group for scenario 1 and 2. As expected, we did not find a statistically significant difference in the scores for self-disclosure between the control group (M=0,2004, SD=0,0236) and the experimental group (M=0,1990, SD=0,0311) conditions; t (26)= -.234, p = .817 for scenario 1.

For scenario 2 another one sample t-test was conducted. Again, we did not find a statistically significant difference in the scores for self-disclosure between the control group (M=0,3143, SD=0,0652) and the experimental group (M=0,3120, SD=0,0677) conditions; t (24)= -.167, p = .869. These results show that there is no statistically significant difference in group means between the control and experimental group for scenario one and two; design similar with and without pre-fill.

(29)

29

Scenario 3: Design not similar, not pre-filled

To test the effect of the manipulation which was applied to the experimental group of scenario three, a one-sample t-test was conducted to compare the probability of self-disclosure between the control and experimental groups. To summarize:

• Control group: Design not similar, not pre-filled • Experimental group: Design similar, not pre-filled

We found a statistically significant difference in the scores for self-disclosure between the control group (M=0,0711, SD=0,0254) and the experimental group (M=0,1798, SD=0,0537) conditions; t (33)=11,80, p = .000. These results show that there is a statistically significant difference in group means between the control and experimental group. The level of self-disclosure in the experimental group with a similar design is significantly higher than the level of self-disclosure in the control group.

Scenario 4: Design not similar, pre-filled

As the manipulation was also applied to scenario four, another one-sample t-test was conducted to compare the probability of self-disclosure between the control and experimental groups. To summarize:

• Control group: Design not similar, pre-filled • Experimental group: Design similar, pre-filled

(30)

30 As hypothesized, we found a statistically significant difference in the scores for self-disclosure between the control group (M=0,1339, SD=0,0178) and the experimental group (M=0,3016, SD=0,0852) conditions; t (25) = 10,04, p = .000. These results show that there is a statistically significant difference in group means between the control and experimental group. Again, the level of self-disclosure in the experimental group with a similar design is significantly higher than the level of self-disclosure in the control group. Based on the findings in scenario three and four, H3a is supported. Switching the design of the customer journey from not similar to similar has a statistically significant and positive effect on the consumers’ willingness of self-disclosure.

5. STUDY B: RESEARCH METHODOLOGY

We conducted an empirical study to gain more knowledge about what factors influence the consumers’ willingness of self-disclosure in a case of online advertising. The first section outlines the sample and the main characteristics of the data. In the second part we go deeper into each of the used constructs and conclude with the analysis methodology.

(31)

31

5.1 Research Design

To test our hypothesis a factorial between-subjects experimental design was used. We collected data by means of an online vignette study followed by an online questionnaire. The experiment contained the same four scenarios as described in study A (design similar & not pre-filled, design similar & filled, design not similar & not filled and design not similar & pre-filled). Each of the participants was randomly assigned to one of the four different scenarios. After the introduction of the survey was presented we asked the participant to reveal gender, age and level of education. After the purpose of the advertising platform and its customer journey were explained, the participant was asked if they would provide their personal information (self-disclosure) to the online advertising platform in order to get the exclusive discount voucher of their choice. Afterwards 34 questions were presented to link the dependent variable self-disclosure to the theory-based constructs as presented in the conceptual

framework. The participants could complete the survey using the Qualtrics4_{survey tool. Before}

the actual survey was sent out, we conducted pretests with students and employees of a company active in online marketing. This led to minor changes in the design and wording of the questionnaire.

5.2 Experimental Design

The online experiment relied on a between-subject design with four different manipulations. At the beginning of the survey, participants received information about the topic of the study, as well as the concept of the online advertising platform used to reward consumers after they have bought something in an online shop. We asked the participants to imagine that they would have bought something in the online shop that was presented in the first visual of the survey. After the order was paid for successfully a check-out page would show up (appendix 9.1) on which a

(32)

32 reward-banner was presented. This reward-banner showed a present box with the text: “A small gift is waiting for you! As a thank you for shopping with us, you can choose an exclusive voucher from our network of partner shops.”. Next, the participant was explained that if they would have clicked on the reward-banner the click would make sure that the participant would get redirected to an external advertising platform on which a selection of exclusive discount vouchers was presented. In scenario one and two the participant was exposed to a voucher page with design similarity (appendix 9.2) which means that the voucher page was designed in the same look and feel of the online shop where the participant initially came from (appendix 9.1). Scenario three and four were designed in a way that the voucher page looked different than the online shop where the order was processed (appendix 9.3). After the consumer was asked to imagine that they would have picked an exclusive discount voucher for one of their favorite shops the pre-filling of personal information manipulation was presented: In scenario one and two the design of the advertising platform remained similar to the online shop but the webform that needs to be filled in to request the discount voucher was either empty in scenario one (appendix 9.4) or already pre-filled with the personal information of the participant in scenario two (appendix 9.5). The same was done for scenario three – ‘not pre-filled’ (appendix 9.6) and scenario four – ‘pre-filled’ (appendix 9.7) in which the design was not similar in comparison to the online shop. Thereafter, we asked the participants to answer questions related to the customer effort score, privacy concerns, personal relevance, trust, brand awareness and the perceived benefits of this service.

5.3 Sample

Non-probability sampling techniques as self- selection, snowball & convenience sampling were used to collect the data that was used for this research. The sample of N=218 participants was recruited via social media websites as Facebook, Twitter and LinkedIn. Additionally, participants were encouraged to share the survey with family and friends. To deal with common

(33)

33 issues related to online data collection we removed cases as follows. First, people that did not complete the survey (n=42) were excluded from the sample. Next, we analyzed the response time and excluded participants who completed the survey in less than 6 minutes (with 11:28 minutes being the average response time). This resulted in a final sample of 218 usable responses out of 271 completed questionnaires.

The mean age was (M= 33.51, SD= 11.72) and 58.7 percent of the participants was male. As 69,7 percent of the participants indicated that they had completed bachelor’s degree or higher, we can state that most of the sample was highly educated. The frequencies of the age and education level can be found in the appendix (§ 9.8).

5.4 Measures

The measures of our independent variables used in this study are mostly taken from existing literature. An extensive overview of all items used to measure the constructs presented below is given in table 2 (appendix 9.9). Unless indicated differently, all items were rated on a 7-point Likert scale.

Self-Disclosure (SD)

The dependent variable self-disclosure (M=0,37, SD=0,48) is measured using closed question which could only be answered with yes/no. The dependent variable is therefore binary. The question was adopted from the research done by Krafft, Arden, & Verhoef (2017). Depending on if the pre-fill condition was applied to the scenario the wording of the question was adjusted.

Customer Effort Score (CES)

The variable customer effort score is measured using a scale adopted from Dabholkar (1994). This scale was measured using three different questions (‘The process of requesting the voucher and filling in my personal details will: be complicated, take a long time for me, will take a lot

(34)

34 of effort for me, 1= strongly disagree to 7= strongly agree). The scale showed reliable (M=2,63, SD= 1,53, α= .86).

Privacy Concerns (PC)

As the independent variable privacy concerns played a key role in this research we have chosen to adapt different scales measuring privacy concerns from Lwin, Wirtz, and Williams (2007) and Dinev & Hart (2006). Reliability analysis showed the scale to be reliable (M= 4,68, SD= 1,50, α= .94).

Personal Relevance (PR)

Personal relevance was measured using scales adopted from Zaichkowsky (1985) and Srinivasan, Anderson and Ponnavolu (2002). The wording of the questions was modified to make them fit in the context of this research. The measurement of this construct used eight items, such as “The discount voucher - will be supposedly relevant to my needs, will be supposedly meaningful to me, will be supposedly useful to me”. Reliability analysis showed the scale to be reliable (M = 4,11, SD = 1,19, α= .89).

Trust (T)

To measure the construct trust we adopted three items from the scale from Dinev & Hart (2006). In addition, three new items were introduced to specifically measure the trustworthiness of the advertising/voucher platform. Statements such as “I trust the company responsible for the voucher platform” and “I trust the voucher platform because this service is offered by this particular online shop (MyJewellery)” were added to the questionnaire. Reliability analysis showed the scale to be reliable (M = 3,52, SD = 1,21, α= .89).

(35)

35

Brand Awareness (BA)

The variable brand awareness is measured using a scale adopted from Oh (2000). This scale was measured using five different statements (‘The brand name MyJewellery is: familiar, known, visible, heard a lot, recognized). The scale showed reliable (M=3,68, SD= 1,41, α= .88).

Perceived Benefits (PB)

Prior survey items were adapted to measure the perceived benefits variable (Xu et al., 2010; Forsythe, Liu, Shannon & Gardner, 2006) with minor modifications in terms of wording to make the scale fit into the context of this research. Reliability analysis showed the scale to be reliable (M = 4,07, SD = 1,44, α= .88).

Control variables:

In this research the control variables used by Awad & Krishnan (2006) were used to gain more insight about the influence of consumer demographics such as gender, age and level of education. Furthermore, in the survey participants were asked if they would click on the reward-banner as soon as they have landed on the check-out page. A control variable named ‘click-rate' was included as consumers can only request a discount voucher in the live environment of the advertising platform if they would have clicked on the reward-banner. However, the survey was designed in a way that it could be that a consumer would not click on the reward-banner and ultimately decide to self-disclose to get the voucher of their choice.

6. RESULTS

6.1 Data Preparation

As we already discussed above (§ 5.3), we have removed 53 cases from the dataset due to the fact that those participants had either not competed the entire survey (n=42) or because the

(36)

36 response time was below 6 minutes (n=11). A dummy variable was created by coding each of the four scenarios as 1, 2, 3 and 4. The dummy variable is named as ‘design similarity & pre-fill’ and is taken into our correlation matrix to get a first impression of the effect on our dependent variable self-disclosure. To check the reliability of the theory-based constructs a reliability analysis was performed. A correlation matrix including the means, standard deviations and results of the reliability tests is presented in table 3.

Table 3: Means, standard deviations, correlations and reliability scores in brackets.

6.2 Data Manipulation Checks

Before we conducted our main analysis a manipulation check was performed to get a first impression of the potential effects of design similarity and pre-filling of personal information on the probability of self-disclosure. At the end of the survey we asked the participants if they would click on the reward-banner (again) now they know what is coming. Since clicking on a banner is a binary decision we asked the participants a closed question: “Would you click on the reward-banner (again) next time?” (0 =”no" and 1 = “yes"). The results of the manipulation check to verify group-level effects is provided in figure 4. The expected click-rate is higher for the scenarios in which the design of the advertising platform is similar to the online shop. Besides, the dummy variable ‘design similarity & pre-fill’ which is representing our

(37)

37 manipulations in the experiment showed that 9 out of 11 Pearson correlations are statistically significant of which 7 at the 0.01 level.

Figure 4. Manipulation check

6.3 Data Analysis

Hypotheses H1 – H6 were tested using SPSS through a linear regression analysis. All analyses were examined with click-rate, age, gender, education and design similarity & pre-fill as control variables. A significant regression equation was found (F (11, 206) = 25.347, p < 0.001), with

an R2 of .575. As none of the control variables showed to be statistically significant another

simple linear regression analysis was conducted leaving out the control variables. Again a

significant regression equation was found (F (6, 211) = 45.627, p < 0.001), with an R2 of .565.

(38)

38 the more concise model to get a clearer picture of the explanatory power of our theory-based constructs on the dependent variable self-disclosure. The regression model used for our main analysis is presented in figure 5. The regression model including the control variables can be found in the appendix (§ 9.10).

To test hypothesis H7 a one-sample t-test was conducted to compare the level of self-disclosure between the different scenarios. Finally, H8 and H9 were tested by the interpretation of a means plot in which the probability of self-disclosure was visualized per scenario.

Figure 5. Linear regression model

6.4 Main Analysis

Many researchers found that privacy concerns are closely related to the consumers’ reluctance of sharing personal data (e.g. Krafft et al., 2017; Norberg et al., 2007; Taddicken, 2014). Results of the Pearson correlation indicated that there was a moderate and significant negative association between privacy concerns and self-disclosure, (r(216) = -.594, p = < .001). As hypothesized, privacy concerns (b = -.088, p < .01) reveals a highly significant negative effect on self-disclosure. In other words, if the level of privacy concerns increases for one, the probability of self-disclosure will decrease for -0.088. Thus, H1 is supported. The higher the perceived privacy concerns, the lower the probability that consumers will self-disclose.

(39)

39 Before the linear regression was conducted the first impression from the Pearson correlation was that trust would have a significant effect on self-disclosure as this construct has the second strongest (moderate) positive correlation, (r(216) = .563, p = < .001). However, although trust could have a positive effect on self-disclosure our results do not statistically

confirm H2, stating that the higher the level of trust, the higher the probability that consumers

will self-disclose (b = .047, p > .05). Thus, H2 is not confirmed in a statistically significant manner.

Prior research showed that the perceived effort during the process of self-disclosure was expected to have a negative effect on consumers’ willingness of self-disclosure (Dickinger, Haghirian, Murphy, & Scharl, 2004; Krishnamurthy, 2001). The Pearson correlation revealed a weak and significant negative correlation between the customer effort score and self-disclosure r(216) = -.296, p = < .001). The regression model also showed a statistically significant and negative association with our dependent variable self-disclosure. If the customer effort score increases for one, the probability of self-disclosure will decrease for -0.034. Hence, our results confirm H3 supporting that the higher the perceived level of effort, the lower the probability that consumers will self-disclose (b = -.034, p < .05).

Following our literature review we expected personal relevance and brand awareness to explain a good amount of the total variance of self-disclosure (Baek & Morimoto, 2012; Lemon & Verhoef, 2016; Zhao, Lu, & Gupta, 2012). Interpreting the Pearson correlations, personal relevance (r(216) = .509, p = < .001) and brand awareness (r(216) = .428, p = < .001) showed to have moderate and positive significant correlations with self-disclosure. Interestingly, the effects of personal relevance (b = .041, p > .05) are positive but not significant. This finding implies that that personal relevance does not directly affect the probability of self-disclosure. Thus, H4 is not confirmed in a statistically significant manner. Similarly to our H4, brand

(40)

40 awareness is positively related to self-disclosure but is not statistically significant (b = .032, p > .05). Thus, H5 is not confirmed.

With different theories stating that cost-benefit trade-offs are the basic principle when it comes to sharing personal information (Awad & Krishnan, 2006; Keith, Thompson, Hale, Lowry, & Greer, 2013) in an online environment we expected the perceived benefits to explain a significant amount of variance of self-disclosure. The Pearson correlation indicated a relative strong effect as our independent variable perceived benefits showed to have the strongest positive correlation of all our variables that were taken into our analysis. The Pearson correlation revealed a moderate and significant positive correlation between the perceived benefits and self-disclosure, r(216) = .665, p = < .001). Interpreting our linear regression model perceived benefits show the most pronounced positive impact on the probability that consumers decide to self-disclose, as indicated by its b value (b = .108, p < .01). This indicates that when participants perceive higher benefits, they will be more inclined to share their personal information. Confirming our H6, the higher the level of perceived benefits, the higher the probability that consumers will self-disclose.

To test our H7 a one-sample t-test was conducted. We hypothesized that a customer journey with a similar design would show a higher probability of self-disclosure in comparison to a customer journey which is not similar. We compared the probability of self-disclosure between scenarios one and two (design similar) versus scenarios three and four (design not similar). Below the descriptive statistics related to the self-disclosure averages and the results of the one-sample t-test are provided:

(41)

41 We found a statistically significant difference in the scores for self-disclosure between the design similar group (M=0,4712, SD=0,5016) and the design not similar group (M=0,2719, SD=0,4469) conditions; t (103) = 4,051, p = .000. These results show that the probability of self-disclosure in the group with a similar design is significantly higher than the probability of self-disclosure in the group design not similar. Thus, H7 is supported. A visual presentation of group means is provided in figure 5.

Figure 5. Means of self-disclosure for design similarity

To test H8 & H9 the probability of self-disclosure per scenario was visualized in a means plot which can be found in figure 6. As hypothesized, scenario three (design not similar, not

(42)

pre-42 filled) does show the lowest probability of self-disclosure. Furthermore, and as expected following the results from study A, we have found that scenario 2 (design similar, pre-filled) does show the highest probability of self-disclosure. Thus, confirming H8 and H9.

Figure 6. Means of self-disclosure per scenario

6.5 Additional Exploratory Analysis

Running the same linear regression model for each of the individual scenarios we found that trust was found to be statistically significant (b = .169, p < .05) in scenario three; design not similar, not pre-filled. This could mean that when consumers have more confidence in the internet and specifically this advertising platform, the probability increases that someone who is visiting an unfamiliar online advertising platform decides to self-disclose to get a discount on a future purchase. However, trust was not found to be statistically significant in the other

A quantitative study on self-disclosure in online advertising