Do fair players win? : the influence of consumers’ fairness evaluation on information disclosure

(1)

M

ASTER

T

HESIS

DO FAIR PLAYERS WIN?

THE INFLUENCE OF CONSUMERS’ FAIRNESS EVALUATION ON INFORMATION

DISCLOSURE

M.Sc. Business Administration | Digital Business | Supervisor: Sara Valentini Katharina Dassel | 11834412 | 22.06.2018

(2)

STATEMENT OF ORIGINALITY

This document is written by Student Katharina Dassel 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)

TABLE OF CONTENT

1 INTRODUCTION 1

2 THEORETICAL FRAMEWORK AND HYPOTHESES 4

2.1 INFORMATION PRIVACY CONCERNS 4

2.2 WILLINGNESS TO DISCLOSE INFORMATION 6

2.3 THEORIES OF JUSTICE 9 2.4 DISTRIBUTIVE JUSTICE 10 2.5 PROCEDURAL JUSTICE 12 3 METHODOLOGY 14 3.1 RESEARCH DESIGN 14 3.2 PRETEST 15 3.3 MEASUREMENT 16 3.4 DATA 17 4 RESULTS 19 5 DISCUSSION 28 5.1 KEY FINDINGS 28

5.2 CONTRIBUTIONS TO THEORY AND PRACTICE 32

5.3 LIMITATIONS AND FUTURE RESEARCH 34

6 CONCLUSION 35

REFERENCES 37

APPENDIX 44

A. ORIGINAL SCENARIO DESCRIPTIONS 44

B. PRE-TEST RESULTS 47

C. ANOVA ANALYSIS FOR PERCEIVED REALISM 48

D. FACTOR ANALYSIS OF THE INDEPENDENT VARIABLES 49

(4)

LIST OF TABLES

TABLE 1: MEASUREMENT INSTRUMENTS ________________________________________ 18

TABLE 2: SAMPLE CHARACTERISTICS AND HOMOGENEITY TEST BETWEEN GROUPS ________ 19

TABLE 3: ITEMS AND RELIABILITY OF THE ITEMS OF WILLINGNESS TO DISCLOSE INFORMATION

_________________________________________________________________________ 21 TABLE 4: MEAN AND STANDARD DEVIATION FOR THE DEPENDENT VARIABLES ACROSS THE

CONDITIONS _______________________________________________________________ 22 TABLE 5: ANOVA RESULTS FOR THE DEPENDENT VARIABLES _______________________ 23

TABLE 6: REGRESSION ANALYSIS RESULTS FOR TESTS OF H2 AND H3 __________________ 24

TABLE 7: REGRESSION ANALYSIS RESULTS FOR TEST OF H2.2 ________________________ 25

TABLE 8: REGRESSION ANALYSIS RESULTS FOR TESTS OF H2.1 AND H3.1 _______________ 25

TABLE 9: RESULTS FOR SOBEL TEST STATISTIC ___________________________________ 26

TABLE 10: FURTHER ANALYSIS OF DEPENDENT VARIABLES __________________________ 27

TABLE 11: RESULTS AND SUMMARY OF HYPOTHESES TESTING _______________________ 27

TABLE 12: MEAN AND STANDARD DEVIATION FOR PRE-TEST RESULTS OF PERCEIVED

MANIPULATIONS ___________________________________________________________ 47 TABLE 13: RESULTS FOR INDEPENDENT SAMPLES T-TEST ____________________________ 47

TABLE 14: MEAN AND STANDARD DEVIATION FOR PRE-TEST RESULTS OF PERCEIVED REALISM 47

TABLE 15: RESULTS OF ONE-WAY ANOVA FOR PRE-TEST ___________________________ 48

TABLE 16: MEAN AND STANDARD DEVIATION FOR THE ITEM OF PERCEIVED REALISM ______ 48

TABLE 17: RESULTS OF ONE-WAY ANOVAS _____________________________________ 48

TABLE 18: FACTOR LOADINGS FOR THE CLUSTERING OF THE INDEPENDENT VARIABLES _____ 49

LIST OF FIGURES

FIGURE 1: CONCEPTUAL FRAMEWORK __________________________________________ 14

(5)

ABSTRACT

Collecting personal information online is essential stay competitive in today’s data-driven world from a company and marketer perspective, but often leads to privacy concerns from a consumer perspective. Despite much attention, there exists contradiction in research around the concept of consumer behavior online, especially with regards to privacy concerns and willingness to disclose information. This study investigates the role of monetary incentive and transparency of the communication on the perceptions of fairness and willingness to disclose personal information online. The theoretical framework is anchored to the theory of perceived justice and consumers’ fairness evaluations. This framework was tested by conducting a scenario-based online experiment, and the data of 265 final responses was analyzed using multiple regression analyses. Results revealed that providing a monetary incentive effectively increases the perceived fairness of the outcome of an information exchange, which indirectly leads to a higher willingness to disclose information. The same holds true for a transparent privacy handling; it increases the perceived fairness of the procedure and leads to a higher willingness to share data. Moreover, the study finds that a transparent handling also reduces privacy concerns. Interestingly, results revealed that privacy concerns do not predict willingness to disclose information online. These results are in line with findings on the privacy paradox and the concept of privacy apathy, showing that people do state their concerns about privacy, but these concerns do not influence behavioral intention. Rather, consumers seem to evaluate their outcome and the privacy handling, and reward those that play fair with the provision of personal information.

(6)

1 Introduction

Data and information privacy has been an issue for consumers for decades (Culnan and Armstrong, 1999). It is no secret that consumer data is the oil of our “data economy” era, the more companies gather to use, the more significant their competitive advantage can be (Economist, 2017). Especially in marketing and related disciplines, it is crucial to make use of consumer data to create customer value and stay competitive. Therefore, marketing practices are advancing in a rapid pace, fueled by technological developments (Martin and Murphy, 2017). These developments though stress the debate about information privacy and not only from a consumer perspective, privacy is becoming more and more a concern (Bélanger and Crossler, 2011). At the same time, consumers are mostly still left in the dark and are often not aware about what data is collected, who has access to their data or how the data is used and feel a growing anxiety about privacy practices applied by firms (Morey, Forbath and Schoop, 2015; Martin and Murphy, 2017). Data breaches and intrusive marketing practices (i.e., at Target or Facebook) heat up the discussion and show that regulation and legislation has its flaws (Duhigg, 2012; Hill, 2012). The call for a changed privacy behavior is coming from scholars and practitioners alike; to (re)gain consumers trust and stay competitive in the game of big data, transparency is considered to be the new communication tool. On top, companies slowly start to realize that handling their customers’ data right can bring not only short-term results (i.e., more information disclosure) but also trust and loyalty (Morey, Forbath and Schoop 2015; Martin et al. 2017; Martin and Murphy, 2017).

One example to consider is the TV channel “Channel4”: in a video on their website, they explain simply and funny why they need their customer’s data and what they are doing with it (Channel4 Website, 2018). With this “Viewer Promise,” they hope to gain consumers understanding for the data collection and want to gather even more data. They are not alone

(7)

with this approach (Morey, Forbath, and Schoop, 2017). Also, Apple communicates privacy as (marketing) strategy by stating for example, that the company refuses to grant U.S law a backdoor to their customer’s phones (Lichtblau and Apuzzo, 2016).

There is a large body of literature on privacy, and different streams of research from different disciplines try to explain privacy concerns, consumer behavior and privacy, or the so-called “privacy paradox”, where privacy behavior is not in line with earlier stated concern (Martin and Murphy, 2017; Kokolakis, 2017). However, none of the existent work in the field of privacy investigates how consumers react to a very transparent privacy handling and information exchange, combined with a monetary compensation for information, which is becoming more and more popular among practitioners. In this study, I empirically investigate how consumers react when a company is transparently handling data privacy matters, or when they are provided with a monetary incentive and are thus explicitly told about the benefits of the information exchange.

Based on the theories of justice and drawing on the dimensions of distributive and procedural justice I propose that when a monetary compensation is offered for information and when consumers are informed transparently and explicitly about the privacy handling, they perceive the company’s privacy practices as fairer and are thus more willing to disclose information, as privacy concerns are reduced.

With this study, I aim to contribute to literature in several ways. Firstly, I expand the literature around theories of justice, trying to unite the understanding of the dimensions of justice related to (information) privacy concerns and the willingness to disclose information in a conceptual model and provide empirical proof for the relationships.

Secondly, I aim to take into consideration the contradictory research on monetary incentives and link the practices of providing a monetary compensation for personal information and a transparent privacy handling as antecedents of the dimensions of distributive and procedural justice.

(8)

Thirdly, the call for more research in the field of privacy and (marketing) strategy has been made several times. The MSI highlights the importance on the topic of the tradeoff between privacy concerns and benefits of personalization and also states “establishing an optimal social contract with customers” as one of their 11 research priorities (Marketing Science Institute, 2016; Martin and Murphy, 2017; Krafft et al., 2017). This research can thus contribute to a new direction in marketing strategy.

To practitioners, the findings of this study can be of enormous value as the answer to the research question can reveal a strategy how companies should treat privacy matters to mitigate privacy concerns and thus gain a competitive advantage. Furthermore, I hope to empirically validate building blocks of essential aspects of future privacy handling, that help firms to formulate a sound privacy strategy.

The study is structured as follows: Section two describes the theoretical framework and the hypotheses. Section three explains the procedure of this research as well as the methodology. Next, the hypotheses are tested, and the results are displayed in section four, while section five gives an overview of the discussion. After that, the section concludes with limitations, future research avenues and closes with a conclusion in part six.

(9)

2 Theoretical Framework and Hypotheses

The issue of privacy in digital environments has found attention in academic research for over two decades. More and more research takes the chance on explaining fundamental principles and drivers that are underlying consumer behavior in digital environments when it comes to privacy. In this chapter, I want to give an overview of the existing literature around the issue of privacy concerns and consumer behavior. First, the research around the influence of privacy concerns on behavior, especially willingness to disclose information, is discussed. Next, the theories of justice and specifically the dimensions of distributive and procedural justice are consulted to explain privacy concerns and its antecedents. This theoretical framework and the derived hypothesis will lay the ground for the conceptual model used in this study.

2.1 Information Privacy Concerns

The concept of consumer privacy is hard to define in literature (Martin and Murphy, 2017). The first definitions referred to in current scholarship date back to Warren and Brandeis (1890), who define privacy as “the right to be left alone.” Westin (1967) uses this and defines privacy as the “right to determine when, how, and to what extent information about ourselves is communicated to others.” Finally, Goodwin (1991, p. 152) provides an overview and definition of consumer privacy as information privacy. She states that information privacy is “control over information disclosure and unwanted intrusions into the consumer’s environment.” Information privacy is also the term most of academic literature refers to in the digital environment. In the relevant context of marketing though, consumer information privacy is described by Foxman and Kilcoyne (1993) and then termed to the digital marketing environment by Nill and Aalberts (2014), both describing the organization’s access, use, dissemination and protection of consumer data for marketing purposes.

(10)

Various streams of research tackle the topic of information privacy from different perspectives, using numerous theoretical frameworks to explain data privacy concerns and consumer attitudes, as well as behavioral outcomes (such as willingness to disclose information). These include social contract theory (Phelps, Nowak and Ferrell, 2000; Gabisch and Milne, 2013), the privacy calculus and justice theory (Culnan and Bies, 2003; Ashworth and Free, 2006; Son and Kim 2008), social exchange theory (Schumann et al 2014; Krafft et al., 2017) and behavioral decision theory (Mothersbaugh, 2012; Acquisti et al., 2013).

A prominent construct of discussion in existing research about privacy evolves around privacy concerns. Privacy concerns have been studied much in offline environments (e.g., Smith et al., 1996) and have been the center of attention, also in the context of marketing (Phelps, Nowak and Ferrell, 2000). Due to long research in silos, there exist different conceptualizations and measurements of privacy concerns (Li, 2011). Smith (1996), stated that privacy concerns arise under four conditions: (1) information collection, (2) unauthorized (internal and external) secondary use, (3) improper access, and (4) error protection. Following research builds on these pillars. A second interesting and well-used concept is called Concerns for Information Privacy (CFIP), later remodeled to the online environment (IUCFIP) (Malhotra et al., 2004). Marketing literature prefers to measure privacy concerns through more direct questions, while recent Information Systems Research has operationalized concerns in the context of online environments more focused on a scale. (Martin and Murphy, 2017, Smith and Dinev, 2011). Malhotra (2004, p. 337) define privacy concerns as “individuals’ subjective views of fairness within the context of information privacy,” and for the purpose of this study, this definition is used. Here, subjective means that concerns are based on personal perceptions and may be different for every person. Another important factor is that privacy concerns are found to be not only subjective but also very context dependent, ranging from extreme concerns to complete privacy apathy (Choi, Park and Jung, 2018). In online environments, privacy concerns generally describe the concern of improper use, use without consent, no transparency

(11)

and control. In the context of consumer and information privacy concerns, the focal area of interest arches from antecedents of concerns to (behavioral) outcomes. This is summarized in the APCO framework (Antecedents – Privacy Concerns – Outcomes) by Smith and Dinev (2011). Although parts of the framework are still studied in silos, it is applicable to most relevant streams of research including the fields of marketing, IS or business ethics. Dinev and Hart (2006) already built a similar framework and found privacy concerns to partially mediate the relationship between perceived risks and willingness to disclose information. Privacy concerns as a mediator and predictor variable towards consumer behavior is supported by literature and is often used to explain consumer privacy outcomes, especially willingness to disclose information as a dependent variable (Milne et al., 2004; Martin and Murphy, 2017; Nam et al., 2006).

2.2 Willingness to disclose information

In many studies, privacy concerns are seen as the main driver to explain disclosure behavior online. In the age of information, entities depend on the disclosure of data, knowingly or unknowingly provided from the consumer, who in turn faces the challenge of navigating his privacy online (Acquisti, Brandimarte, and Loewnstein, 2015). To stay in the scope of this study, the dependent variable can be defined as “the personal information individuals intentionally and voluntarily reveal about themselves to others in an interpersonal relationship” (Li 2012, p. 166), applied to an online registration process.

Not only in marketing practices, willingness to disclose information still has an immense impact on understanding the behavioral outcome of privacy concerns (Zhang Liu, 2017). The theory of reasoned action is highly cited in this context (Fishbein and Ajzen, 2010), together with the expanded theory of planned behavior (Ajzen, 1991), suggesting that attitudes influence intentions, which in turn directly affect actual behavior. In the realms of information privacy ambiguous findings of the privacy paradox, where stated attitudes and intentions do not

(12)

match behavior, raise the question if this theory is still valid (Norberg, 2007). Other streams of research can still find a causal relationship between intention and behavior and suggest that different contextual factors can be blamed for the discrepancy in the privacy paradox (Choi and Jung, 2018).

Several theories look at intention to disclose information as a cognitive process and different fundamental theories are applied by scholars to explain and understand consumer behavior in the context of privacy concerns (Li, 2012). Culnan and Bies (2003) draw on the concept of social exchange theory and theories of justice, stating that perceived fairness is critical to fulfill a social contract in an information exchange relationship. Information collection is seen as a marketing exchange relationship (Martin and Murphy, 2017) and in line with the privacy calculus, it is argued that information will be disclosed if perceived benefits of the exchange outweigh perceived risks (Culnan and Bies, 2003). If risks, that means privacy concerns, outweigh the perceived benefits, the exchange is perceived as unfair and not only the intention to disclose information goes down but also negative responses as falsifying information, complaining or negative word of mouth can occur (Son and Kim, 2008). To avoid negative responses like that it is essential to consider the justice perceptions and privacy concerns as a relevant driver for disclosure intention (Norberg et al., 2007).

It has been found and supported by different researchers that privacy concerns are negatively related to willingness to disclose information (Dinev and Hart, 2006; Ward and Bridges, 2007). Clearly, it can be stated, that also in online environments it is true that the lower the privacy concerns, the higher the willingness to disclose personal information (Wakefield, 2013 and Yang, 2009). In line with existing research I therefore also argue for this study, that privacy concerns have a negative impact on willingness to disclose personal information in a website registration process online. Hence:

(13)

Intention to disclose information and thus disclosure behavior is a multidimensional construct (Knijnenburg, Kobsa, and Jin, 2013). Concerns are highly context- dependent (Acquisti, Brandimarte, and Loewnstein, 2015) and concerns and willingness to disclose differ for different types of requested data (Mothersbaugh, 2012). Although Phelps, Nowak, and Ferrell (2000) examined five broad types of different data, there has been limited research around how the consumer classifies different types of information based on concerns and risk perceptions. Nowadays it is also possible to turn former classified as unidentifiable information into identifiers, thanks to technological advancements (Schwarz and Solove, 2011). Milne et al. (2017) therefore empirically propose a new typology of information types, which is organized by similar risk perceptions of the consumer: basic demographics, personal preferences, contact information, community interaction, financial information and secure identifiers. The authors go a step further than Phelps, Nowak, and Ferrell (2000), who find that willingness to disclose information differs for different kinds of information, by integrating the factor of different perceived risks by consumers. Therefore they acknowledge concerns as a mediating force. Milne et al.’s (2017) cluster of basic demographics include among others birth date, gender, and shopping behavior. They found this to be the cluster with the lowest risk perception which is in line with the findings of Phelps Nowak and Ferrell (2000) who found consumers to be most likely to provide demographic information. Data like e-mail address, weight or sexual preference are clustered under personal preferences and have the second lowest risk perception (Mile et al., 2017). Contact information such as mobile phone number and home address are perceived riskier in comparison, which is also in line with a lower willingness to provide, found from Phelps Nowak and Ferrell (2000).

Interestingly, community interaction, which includes the social network profile and family and friend’s contact information, is rated with a higher risk. As this study looks at an information exchange during a registration process, this type of information is highly relevant to consider. At the same time, financial information and secure identifiers like social security

(14)

number are irrelevant to include. It will be interesting to see how the request of different types of data leads to different concerns and how it influences the willingness to provide this information.

Approaching the problem of information privacy by merely taking the outcome of the privacy calculus into account has its limitations though. Considering only concerns and outcomes neglects to draw the connection to the antecedents of privacy concerns, which play a crucial role in understanding consumer behavior (Zhan, Liu, and Cheng, 2017, Smith and Dinev, 2011). It is common to use privacy concerns as a predictor variable towards behavioral outcomes, but under more complex conditions, which are closer in portraying reality, privacy concerns are conceptualized as a mediating condition (Martin and Murphy, 2017; Bleier and Eisenbeiss, 2015). Accordingly, the theories of justice are consulted to shed light on the relationship between antecedents, privacy concerns, and outcomes.

2.3 Theories of Justice

Theories of justice have been widely applied to understand behavior in different streams of literature, including organizational theory and service literature. Specifically, in information privacy literature, it has become a popular approach (Xu Teo and Tan, 2009). The framework is considered a fruitful area to explain consumer behavior related to privacy concerns in online as well as offline environments (Culnan and Bies, 2003; Ashworth and Free, 2006; Son and Kim, 2008; Martin and Murphy, 2017).

Research suggests that consumers evaluate privacy handling as fair or unfair. Therefore, fairness is perceived as a critical component and antecedent of information privacy concerns (Ashworth and Free, 2006). Fairness perceptions also explain contextual constraints (Martin and Murphy, 2017), specifically perceived justice is thought to lead to lower privacy concerns and higher willingness to disclose information (Lanier and Saini, 2008; Culnan and Bies, 2003).

(15)

Scholars do not agree on the dimensionality of the construct of theories of justice (Ashworth and Free, 2006, Xu, Teo, and Tan, 2009). Two up to five dimensions are described in different streams of literature, but it can be concluded, that all these are reflected by the two main dimensions: distributive and procedural justice (Colquitt, 2001; Xu Teo, and Tan, 2009). In contexts where the construct is used to explain privacy concerns and willingness to disclose information in online environments, it is argued for the dimensions of distributive and procedural justice (Ashworth and Free, 2006). Thus, this study focuses on these two dimensions, which are disclosed in the next sections.

2.4 Distributive Justice

Distributive justice, initially introduced by Homans (1961) as the perceived fairness of the outcomes one receives in an exchange relationship, is in more recent literature specified and referred to as the exchange of benefits consumers receive in an online exchange for the release of their personal information (Martin and Muprhy, 2017; Son and Kim, 2008). It is argued, that when evaluating the exchange, a form of the privacy calculus is applied (Dinev and Hart, 2006). Consumers are aware of the monetary value of data (Krafft, 2017; Tsai, 2011) and data disclosure is perceived as a personal sacrifice (Son and Kim, 2008). To evaluate the outcome of the exchange consequently as fair, the consumer expects an equivalent which is comparative to some standard (Xu, Teo, and Tan, 2009). The fairness is therefore judged based on the proportion of immediate or expected outcome (in the form of monetary incentive or relevant marketing) to input (personal information) for the consumer weighted against the outcome (data) to input (money/effort) proportion of the firm. The more balanced this proportion is, the fairer it is seen. This lowers privacy concerns and increases willingness to disclose information ( "#$%"&'()*+,-./

012#$()*+,-./ =

"#$%"&'

(16)

A typical business practice to ensure that consumers evaluate the outcome as fair is to offer a monetary incentive. This incentive weighs as additional positive outcome for the consumer and therefore influences the fairness perception positively. Scholars support these findings in several ways. Hann, Hui, Lee et al., (2007) showed that incentives increase consumers’ willingness to provide personal information online and that people share their information with fewer privacy concerns (Sheehan and Hoy, 2000). Xi et al. (2006) present the findings that a monetary incentive can even be the single reason why people disclose information.

On the other hand, Lee, Lim, Kim et al. (2015) demonstrate that although monetary rewards are still used in practice to alleviate privacy concerns, they increase concerns and thus lower information disclosure if sensitive information is required. Research about compensation is to some extent contradictory, a recent study also argues that there is no significant relationship between incentives and the decision to grant permission for marketing (Krafft et al., 2017). A different paper states that compensation for information disclosure can even be perceived as negative and manipulating (Pick et al., 2016).

Supported from the underlying mechanism of the evaluation of the dimension of distributive justice and the outcome proportion, this study argues though, that by providing an additional monetary incentive, the outcome will be evaluated as fair, increasing distributive justice. Furthermore, it is argued that distributive justice has an impact on privacy concerns and therefore mediates the effect of monetary compensation on privacy concerns. Thus:

H2: A monetary incentive increases the perceived distributive justice.

H2.1: Distributive justice mediates the effect of monetary compensation on privacy concerns.

Contradictory research regarding incentives states that providing a monetary incentive for personal information can also be evaluated as unfair. This reaction can be explained by psychological reactance, which means that if a feeling of manipulation arises and behavioral

(17)

freedom is perceived to be taken away, this will motivate to restore the original freedom (Brehm, 1966). The freedom to register for a service while providing information (or not) can be restricted for the consumer when an incentive is offered primarily for more sensitive information, leading to a feeling of being manipulated and forced into the registration and information release. This will have an impact on perceived distributive justice as it may rise skepticism and increase concerns. Based on reactance theory it can therefore be argued:

H2.2: Perceived manipulation mitigates the positive influence of an incentive on distributive justice.

2.5 Procedural Justice

Research around the theories of justice indicates that consumers evaluate not only the fairness of outcomes of an information exchange, but also the procedure of how the outcome is reached plays an important role when fairness is assessed.

Procedural justice, as the second dimension is referred to, suggests that the way how people are treated influences the fairness of the procedure evaluated from a consumer perspective (Thibault and Walker, 1975). Central to the perception of fairness has been the influence of awareness (Folger and Bies, 1989), as well as control (Folger and Greenberg, 1985). In the context of online information privacy, these two concepts are used and applied also by Son and Kim (2008), who define procedural justice as the “degree to which an Internet user perceives that online companies give him or her procedures for control of information privacy and make him or her aware of the procedures” (Son and Kim, 2008 p. 508). Ashworth and Free (2006) formulate four normative standards that are also linked to the concepts mentioned above, namely openness, information access, permission, and honesty. The authors argue that consumers compare their treatment to these normative standards for every procedure evaluation. Next to these, theory-driven standards regulation and information principles serve

(18)

as guidelines to ensure such a fair procedure (Xu, Teo, and Tan, 2009). In legislation, specifically the norms information access and permission are legally binding.

The concept of awareness and transparency though is treated negligently. Privacy policies are lengthy, complicated and rarely read at all, although it has been showed that online consumers value it if they are aware of the practices (Malhotra, 2004). The authors also found that the mere existence of privacy policies or additional privacy seals lead people to forget their concerns and disclose more information. Tsai et al. (2001) revealed that when information about privacy is made available it has a positive impact on behavioral outcome. Consumers are even willing to pay a price premium for more accessible information. In the context of consumer privacy, it has been found that when consumers are told that fair information practices are applied, they are more willing to disclose personal information as privacy concerns are mitigated (Culnan and Armstrong, 1999).

This concept of transparency in information privacy research brings up contradictory results when it comes to transparency and privacy concerns particularly. Awad and Krishnan (2006) state that although customers value transparency, once fully educated, they are less willing to be profiled for online for personalization. This research goes in the same direction of John, Acquisti, and Loewenstein (2011), who find that priming the information by a company with a privacy statement (i.e., providing transparency about practices) results in a decrease of information disclosure. Nevertheless, several calls have still been made by practitioners and scholars alike to educate consumers and inform them coherently over the procedure of how privacy is handled, because this is thought to influence the perceived justice positively (Son and Kim, 2008; Awad and Krishnan, 2006; Martin et al., 2017). Theories of justice would argue, that with an increased transparent handling, the procedure of the privacy handling would be perceived as fairer. It is also argued, that procedural justice has an impact on privacy concerns and thus mediates the effect of procedure transparency on privacy concerns.

(19)

H3.1: Procedural Justice mediates the effect of procedure transparency on privacy concern.

The following conceptual framework sums up the hypotheses and proposed relationships.

Figure 1: Conceptual Framework

3 Methodology

3.1 Research Design

For this study an experimental research design with a 2 (with/without monetary incentive) X 2 (high/low transparency), between-subject online survey experiment was chosen and conducted to test the proposed model. This approach allows for the manipulation of personal data request’s attributes, which have an impact on distributive and procedural justice. Distributive and procedural justice are critical antecedents of privacy concern, which in turn mediates the relationship towards willingness to disclose information. The 2X2 experimental design requires four experimental groups. Four scenarios, imitating a registration process for a shopping club online, were developed in line with the manipulation variables. In each scenario, respondents were asked to imagine themselves being in the situation of registering online for a shopping club. To register, they were asked to fill out personal information. Scenario I and II were designed to reflect the high transparency situation; the scenario descriptions were equipped with an additional statement on how the data is collected and how it is used to benefit the consumer,

(20)

assuring full control over personal data as well as referring to the privacy policies. Scenario I and III were manipulated to reflect the monetary compensation condition: for every additional information respondents would provide to the shopping club during the registration, they would get 2€ monetary compensation off their first order. Scenario IV is used as a control condition, providing no monetary incentive and just a standard notification referring to the privacy policies, indicating a low transparent procedure. Respondents were randomly assigned to one of the four experimental groups.

3.2 Pretest

An online pretest was conducted with 49 respondents, to examine the validity and perceived realism of each scenario as well as to assess the clarity of instructions and questions. Respondents reported if they perceived the offering of a monetary incentive as well as the perceived transparency of the privacy handling, to verify the experimental setting of each case. Paired t-tests indicated that the respondents who were presented with a monetary incentive in exchange for personal information (scenario I and III) did perceive this. The mean score of the cases without monetary compensation was 2.58, while the mean of the incentive cases was 5.64 (t= 6.402, p< .05). This difference in means shows that the manipulation works. Cases with high transparency scored significantly different compared to low transparency cases (mean score of high transparency cases m= 5.0; low transparency m= 3.46; t= 3.751, p< .05). Therefore, both manipulations seem to work. For the perceived realism of the study though, the means differed per scenario, indicating that the incentive scenario is not perceived as realistic as the others. These results might be due to the usage of only one item, which is why a richer scale was employed to test realism in the main study again. The final scenarios with four different conditions are shown in Figure 2, the full-text scenarios can be seen in Appendix A, and the results of the pretest are summarized in Appendix B.

(21)

Monetary Incentive No Monetary Incentive Procedure with high

Transparency

Scenario I Scenario II

Compensation: For every

information provided: 2€

Compensation: None Transparency: Notice about how

the data is collected, what it is used for and what the benefits about providing it are, link to

privacy policy

Transparency: Notice about how

the data is collected, what it is used for and what the benefits about providing it are, link to

privacy policy

Procedure with low Transparency

Scenario III Scenario IV

Compensation: For every

information provided: 2€

Compensation: None Transparency: Link to privacy

policy

Transparency: Link to privacy

policy Figure 2: Overview of the final scenarios

3.3 Measurement

Based on the assessment of the presented scenario, respondents were asked to complete an online questionnaire. The measurement items for the questionnaire were adapted from measurement scales of prior studies in the same field to the extent possible. The variables distributive and procedural justice were assessed, using a three-item-scale adapted from literature (Blodgett, Hill, and Tax, (1997); Son and Kim, (2008); Awad and Krishnan (2006)). The measurement scale and items used to measure privacy concerns were adapted from Dinev and Hart (2004), Malhotra (2004) and Knijnenburg, Kobsa and Jin, (2013). From reactance theory literature, one item was measured asking for perceived intrusiveness. For the dependent variable willingness to disclose information respondents were presented with a 7-point Likert scale to assess how likely they would be to reveal the specific information to the shopping club. The information requested was based on the four relevant information typologies discussed by Milne (2017) and in line with Phelps, Nowak, and Ferrell (2000), so that the categories basic demographics, personal preferences, contact information and community interaction were represented. Additionally, a richer scale with four items was used to check for perceived realism. For additional analysis and concerning control variables, respondents were asked to report on their online shopping behavior as well as on deal proneness, while also being asked

(22)

for age, gender, home country and highest obtained degree. The measurement items used in this research are summarized in Table 1.

3.4 Data

The experiment was designed using Qualtrics software, assuring that the randomization of the conditions is working. The first 200 respondents were recruited using Amazon Mechanical Turk. The obtained responses were filtered to eliminate those that had a response time lower than what is needed to accurately read the survey and respond to the questions (min. 4 minutes). Additionally, the survey experiment was distributed through the standard online channels. That resulted in a total of 325 responses that were marked, additionally meaningless responses were eliminated. This elimination resulted in a total of 265 usable responses. For each of the four scenarios, the number of subjects ranged from 65 to 68. The age of participants ranged from 18 to 81, with the majority (51,13%) being between 18 and 30. Most of the respondents came from Germany and the USA (68,4%). 203 respondents (over 75%) reported obtaining a University Bachelor’s degree or higher, indicating that the sample is well educated. Participants who look for offers online at least once per month form 71,13 %, with the majority (85 respondents) indicating to look for offers 1-2 times per week. On top, e-commerce experience and shopping frequency among participants was quite high, with 147 (55,47%) claiming to buy from fashion e-commerce sites more than 1-2 times per 3 months. These results indicate that the sample was able to assess the context of the study of a registration process for an online shopping club. To check for sample differences in the different scenarios, a chi-square homogeneity test was conducted. All groups are homogenous, verifying that the random assignment of participants to the different scenarios was effective. The results are summarized in Table 2.

(23)

Table 1: Measurement Instruments Variable Description

Dependent Variable

Willingness to disclose information

Personal Preferences E-Mail Clothing size

Contact Information Mobile phone number Address

Basic demographics Date of Birth Gender

Community interaction Sign up with Facebook Sign up with google sign up with LinkedIn

(1. Very unlikely ... 7. Very likely) Independent Variable

Monetary Compensation

Considering the scenario, the amounts of benefits I receive in exchange for my information are fair

Given the circumstance of releasing personal information, the incentive for registration offered by deals4fashion is adequate

Giving up my personal information for the amounts of benefits I receive in this scenario is not a fair deal

Reactance I feel manipulated into providing my personal information

(1. Strongly disagree... 7. Strongly agree)

Procedure

Transparency Deals4fashion handles privacy in a very fair manner

Deals4fashion makes reasonable effort to explain how and why my personal information is collected and used

The company's privacy handling is not adequate

After reading the scenario fully, I am more aware of how my data is handled than before

Privacy Concerns I am concerned that the company gathers too much information about me

I am concerned that deals4fashionuses the information for purposes other than the reason I provided it for

I am not concerned that deals4fashion shares my personal information with other parties

Perceived Realism The described scenario of registering for a shopping club is realistic The described scenario of registering for a shopping club is believable

Evaluating which information I share online is something I would also do in real life

(1. Strongly disagree... 7. Strongly agree) Controls

E-Commerce Purchase

Frequency 1. Never; ... 7. Almost every day Looking for deals

online 1. Never; ... 7. Almost every day Age 1. Teenager; ... 5. Over 50 Gender 0. Male; 1. Female Educational

Background 1. High School; 5. PhD Home Country 1. Germany; 5 USA

(24)

Table 2: Sample Characteristics and homogeneity test between groups

4 Results

Realism and Reliability Test

To confirm the experimental validity of the study and the scenarios, respondents were asked to report on the perceived realism of the different experimental conditions. One-way ANOVAs were conducted to observe if there are differences in means for the four items that checked for perceived realism. No significant differences in the means concerning the different scenarios could be observed, indicating that respondents perceived the scenarios as realistic and

(25)

believable. Moreover, respondents reported that they could also identify with the exercise of evaluating which information they share online. The results of the ANOVA are summarized in Appendix C.

Cronbach’s alpha was calculated to test the reliability as there were multiple items employed to measure the constructs of distributive and procedural justice as well as of privacy concerns. When including all items in the reliability analysis Cronbach’s alpha was not satisfying for all three variables, yielding results far below 0.70. For distributive justice, the corrected item-total correlation (with items < .30) and Cronbach’s Alpha if item deleted indicated an item removal of the third item. After revising the content of these items (‘Giving up my personal information for the amount of benefits I receive in this scenario is not a fair deal’), it showed, that it was reversed coded. The resulting removal from further analysis confirmed the scale with the items adapted from literature. This removal led to a reliable Cronbach’s Alpha for the construct of distributive justice of 0.892, with the remaining items having a good correlation with the total score of the scale (all above 0.80). The same procedure was applied to detect reliability of the two other constructs, leading to similar results: for procedural justice as well as for privacy concerns Cronbach’s Alpha was not satisfying when the reverse coded item was included. Removal of these items generated again a reliable Cronbach’s alpha of 0.804 for distributive justice and one of 0.814 for privacy concerns. The resulting scaled-down constructs were chosen to be included for further analysis.

Factor Analysis

A factor analysis has been conducted to identify the underlying constructs measured through multi-item scales of the three independent variables. A Varimax Keizer rotation was applied. The factor analysis identified three well-distinguished factors, that can explain over 82.41% of the variance. The items load highly for the suggested variable, resulting in a factor for distributive justice, procedural justice and privacy concerns each, in line with literature. These factors were saved as variables for further analyses. The last item of procedural justice

(26)

(‘Deals4fashion handles privacy in a very fair manner’) shows cross-loadings on the factor of distributive justice, which could be due to the content and wording of the item. The factor loadings together with the results from reliability analysis can be found in Appendix D.

Dependent Variable: Willingness to Disclose Information

Previous research suggests that the dependent variable can be clustered into four categories, namely basic demographics, personal preferences, contact information and community interaction (Milne et al., 2017), that together form the dependent variable willingness to disclose information. To check the reliability of these clusters as well as to simplify the information for the subsequent analyses, a second factor analysis was done. Here, a three factors extraction is suggested using Varimax Keizer rotation. Interestingly, the loadings corresponded only with minor changes with the proposed clustering. Table 3 shows the factor loadings after rotation, the items that cluster on the same factors suggest that factor 1 represents community interaction as all social-media log in items load highly for it; factor 2 can be described as contact information and factor 3 as personal information. Merging the items in such a way is also relevant as they score similarly in general descriptive results.

Table 3: Items and reliability of the items of Willingness to Disclose Information

Scale and items

Components

1 2 3

Likelihood to reveal - Sign up with LinkedIn .907 Likelihood to reveal - Sign up with google .876 Likelihood to reveal - Sign up with Facebook .831

Likelihood to reveal - Address .349 .783

Likelihood to reveal - E-Mail .729 .393

Likelihood to reveal - Mobile phone number .545 .631

Likelihood to reveal - Date of birth .565 .354

Likelihood to reveal - Clothing Size .905

Likelihood to reveal - Gender .302 .833

Cronbach’s Alpha .870 .690 .673

Notes. Extraction Method: Principal Component Analysis. Rotation Method: Varimax with Kaiser

Normalization.

ANOVA Analysis of Dependent Variables

In order to examine if there are differences across experimental conditions regarding the dependent variables on the basis of the different scenarios, three ANOVAS were conducted.

(27)

One ANOVA used the total mean of community interaction (comm_TOT), the others used the total means of contact information (contact_TOT) or personal information (personal_TOT) as dependent variables. The mean score, as well as the standard deviations, are listed in Table 4. The results indicate that overall, the willingness to disclose community interaction information (total M= 2.95) differs from willingness to disclose contact information (M= 4.34) or personal information (M= 5.37), which scores highest. The mean value for community interaction was below 4.0 (somewhat unwilling to disclose) in all groups, although in groups where an incentive was provided the willingness to disclose this information seemed to be slightly higher.

The same can be observed for the other two dependent variables, the scenarios where an incentive is present (scenario I and III) yield slightly higher results than the ones without. Personal information provision scores highest regarding willingness to disclose, being the only dependent variable with a total overall score > 5.0 (somewhat willing to disclose).

There was a statistically significant difference in conditions for the dependent variable willingness to disclose community interaction information (F (3, 264) = 2.674, p= .048). Tukey post-hoc test revealed that the willingness to disclose was significantly higher in the incentive conditions than in the transparency conditions (p= .03). For contact information, no significant effect could be observed while for personal information, the effect was significant on a 10% level, F (3, 264) = 2.21, p= .087. The results are summarized in Table 5.

Table 4: Mean and Standard Deviation for the Dependent Variables across the conditions

Total Scenario I Scenario II Scenario III Scenario IV Dependent Variable N = 265 n = 65 n = 68 n= 64 n = 68 community interaction 2.95 (1.89) 3.02 (1.91) 2.56 (1.722) 3.45 (2.03) 2.81 (1.84)

contact information 4.34 (1.56) 4.45 (1.61) 4.06 (1.49) 4.54 (1.71) 4.32 (1.38) personal information 5.37 (1.32) 5.6 (1.32) 5.27 (1.05) 5.53 (1.45) 5.09 (1.4)

Notes. Scenario I: incentive x high transparency, Type II: no incentive x high transparency, Type III: incentive x

(28)

Table 5: ANOVA results for the Dependent Variables

Dependent Variable SS DF MS F p- Value

community interaction Between Groups 28.257 3 9.419 2.674 .048 Within groups 929.953 264 3.523 Total 958.21 267 contact information Between Groups 8.857 3 2.952 1.232 .298 Within groups 632.692 264 2.397 Total 641.549 267 personal information Between Groups 11.427 3 3.809 2.21 .087 Within groups 455.089 264 1.724 Total 466.516 267

Notes. SS, sum of squares; DF, degrees of freedom; MS, mean square

Hypotheses testing

Several regression analyses were conducted to test the hypotheses. To follow the logic of the conceptual framework, it was first checked if the manipulation of monetary compensation, on the one hand, is a significant predictor for distributive justice and if procedural transparency, on the other hand, is a significant predictor for procedural justice. Following the argumentation and the framework, it was tested if distributive justice and procedural justice affect privacy concerns, after which a test for mediation was performed in line with H2.1 and H3.1. Hypotheses testing was concluded by examining H1, the relationship of privacy concerns on the three main dependent variables of willingness to disclose information. Further analyses of the relationships between the present independent variables and the three dependent variables complete the results of this study.

To firstly examine if an incentive (IV_incentive) impacts distributive justice on the one hand and if a transparent procedure (IV_transparency) impacts procedural justice, two regression analyses were conducted. In this way, it was made sure that the manipulations predict perceived justice in both dimensions. The results are summarized in Table 6.

Five predictors were entered: age, gender, the transparency condition, the incentive condition as well as the interaction between transparency and incentive condition (incentiveXtransparency). The model was statistically significant, (F (5, 258) = 3.360, p= ,006), and explained 6.1% of variance in distributive justice. The coefficients revealed the incentive

(29)

condition to be the only significant predictor of distributive justice (ß= .252, p= .004). The same was done with procedural justice as dependent variable. Here, the model showed significance with F (5, 258) = 7.041, p= .000. The total variance explained by the model was 12%. In this model only the transparency condition revealed to be a statistically significant predictor variable (ß= .245, p= .003). It can be concluded, that an incentive does predict and increase perceived distributive fairness on the one hand and a transparent procedure does predict and increase perceived procedural fairness on the other, no interaction effect is present. These results indicate that hypotheses 2 and 3 can be supported.

Table 6: Regression Analysis Results for tests of H2 and H3

Variable R R2 R2Change B SE ß t Dependent Variable: Distributive Justice .247 .061 .043 IV_incentive .504 .171 .252** 2.939 IV_transparency -.003 .169 -.001 -.015 incentiveXtransparency -.368 .242 -.159 -1.525 Age -.146 .061 -.146 -1.402 Gender -.03 .122 -.015 -.242 Dependent Variable: Procedural Justice .347 .12 .103 IV_incentive -.217 .166 -.108 -1.305 IV_transparency .49 .164 .245** .299 incentiveXtransparency .33 .234 .142 1.412 Age .084 .059 .084 1.421 Gender .178 .118 .089 1.51

Notes. Statistical Significance: * p < .1; ** p < .05; *** p <.01

H2.2 states that this positive effect can be mitigated if the providing of an incentive is perceived as manipulative. To examine this, a test for moderation was conducted. An interaction variable with the incentive condition and the item that checks for perceived manipulation was created to perform another regression. As summarized in table 7, the model was significant (F (3, 261) = 6.682, p= .000) and the interaction effect showed a significant, negative influence on the dependent variable (ß= -.126, p= .054). Therefore, H2.2 can be supported.

(30)

Table 7: Regression Analysis Results for test of H2.2

Variable R R2 _R2 _Change _B _SE _ß _t

Dependent Variable: Distributive Justice .027 .071 .061

IV_incentive .881 .319 .441*** 2.764

incentiveXreactance -.126 .065 -.327* -1.935

reactance -.037 .045 -.068 -.814

Notes. Statistical Significance: * p < .1; ** p < .05; *** p <.01

To test H2.1 and H3.1 on the mediating role of if distributive justice and procedural justice towards privacy concerns, a Sobel test through a regression was performed. The regression results reveal that distributive justice significantly lowers privacy concerns, while procedural justice does not record an impact on concerns. It is interesting to note that the transparency variable though has a direct effect on privacy concerns, recording a negative ß value of ß= -2.14, p= .014. The same is true for the incentive variable with ß= .178, p= .044. These results, summarized in table 8, indicate that distributive justice partially mediates the relationship between monetary incentive and privacy concerns. Simple regressions were performed, and the unstandardized coefficients, as well as the standard error, were inserted into the Sobel test for mediation. The test statistic reported a p-value of p= .045 for the mediating role of distributive justice. Therefore, H2.1 can be supported. By contrast, H3.1 is not supported, as the non-significant impact of procedural justice on concerns indicates. This is confirmed by the Sobel test, recording a p-value of p= .1. The Sobel test statistic for both mediation test is extracted in Table 9, the coefficients can be examined in Appendix E. Table 8: Regression Analysis Results for tests of H2.1 and H3.1

Variable R R2 _R2_Change _B _SE _ß _t

Dependent Variable: Privacy

Concerns .266 .071 .038 IV_incentive -.356 .176 .178** -2.025 IV_transparency -.429 .173 -.214** -2.479 incentiveXtransparency .465 .245 .201 1.8 distributive -.184 .061 -.184*** -3.035 procedural -.04 .067 -.04 -.594 Age .11 .063 .11 1.758 Gender .146 .124 .073 1.184

(31)

Table 9: Results for Sobel Test Statistic

Variable tested for Mediation Indirect Effect Sobel SE Z-Value p-Value

Distributive Justice -.059 .0298 -2.0006 .0454

Procedural Justice 0 .0389 0 1

To test H1 through a regression analysis, the effect of privacy concerns on the three different factors of willingness to disclose information was examined. Neither for the dependent variable of community interaction nor for contact or personal information a significant model could be established. Thus, H1 has to be rejected and the mediating role of concerns (as suggested by literature) is not taken into account in analysis.

Further Analysis

Another regression was performed to examine if there are effects between the remaining independent variables distributive justice and procedural justice and the dependent variables of willingness to disclose information.

For the dependent variable community interaction, the model was statistically significant (F (9, 254) = 16.2328, p= .000) and explained 36.7% of variance in willingness to disclose community interaction data, namely login with Facebook, google or LinkedIn. Here, four predictors proved to be statistically significant, with procedural justice recording a higher ß-value (ß= .404, p <.001) than distributive justice (ß= .391, p< .001) and gender. For contact information, also a statistically significant model could be built (F (9, 254) = 7.231, p= .000), being able to explain 20.4 % of the variance in this outcome variable. Two predictors were found to have a statistically significant impact; distributive justice showed a direct effect with ß= .401, p < .001 followed by procedural justice with ß= .121, p < .05. Finally, also for personal information the model showed statistical significance, (F (9, 254) = 2.014, p= .038), with 6,7% of the variance explained. The factor of distributive justice is the only predictor of this dependent variable, with ß= .161, p= .01. The results, including all coefficients for the control variables, can be derived from Table 10.

(32)

Table 10: Further Analysis of Dependent Variables Variable R R2 _R2_Change _B _SE _ß _t Dependent variable: Community interaction .605 .367 .344 concern -.001 .051 -.001 -.027 procedural .404 .053 0.404*** 7.575 distributive .391 .052 0.391*** 7.578 Gender 1.931 .827 .065 1.335 IV_incentive .274 .145 .137 1.882 IV_transparency -.255 .144 -.128 -1.774 incentiveXtransparency -.177 .203 -.076 -.874 Age -.006 -.005 -.066 -1.29

Dependent variable: Contact

Information .452 .204 .176 concern .04 .057 .04 .704 procedural .12 .059 .121* 2.022 distributive .401 .057 .404*** 6.996 Gender .121 .92 .614 1.326 IV_incentive -.137 .162 -.069 -.85 IV_transparency -.243 .16 -.123 -1.52 incentiveXtransparency .339 .226 .147 1.503 Age -.009 .005 -.105 -1.845 Dependent variable: Personal Data .258 .067 .034 concern .117 .062 .117 1.884 procedural .053 .065 .053 .826 distributive .161 .063 0.161* 2.579 Gender .043 1.004 .021 .042 IV_incentive .225 .177 .113 1.274 IV_transparency .268 .175 .134 1.533 incentiveXtransparency -.212 .246 -.092 -.862 Age .004 .005 .048 .773

Note. Statistical Significance: * p < .1; ** p < .05; *** p <.01

Finally, Table 11 summarizes the results of the hypothesis including the method of analysis.

Table 11: Results and Summary of Hypotheses Testing

Hypothesis Analysis Method Result

H1: Privacy Concerns ---> Willingness to Disclose Regression Analysis rejected H2: Incentive ---> Distributive Justice Regression Analysis supported H2.1: Distributive Justice ---> Privacy Concerns Regression Analysis supported H2.2: IncentiveXreactance ---> Distributive Justice Test for Moderation supported H3: Procedural Transparency ---> Procedural Justice Regression Analysis supported H3.1: Procedural Justice ---> Privacy Concerns Regression Analysis rejected

(33)

5 Discussion

5.1 Key Findings

The primary purpose of the study was to examine the effect of monetary compensation and procedural transparency on the dimensions of distributive and procedural justice and to empirically explore role of the justice dimensions as influencers on privacy concerns and willingness to disclose information.

The descriptive analysis and the first ANOVA revealed the scores of the three main dependent variables of willingness to disclose information. The results show a significant difference in means for the dependent variables of community interaction and personal information. The reason for this difference in means is the monetary compensation that was offered for additional information.

It stands out that participants were in general not very likely to disclose especially their social media log-in (community interaction), only concerning contact information the tendency goes towards respondents being somewhat likely to disclose. This is in line with the clusters and risk perceptions about disclosure. Milne et al. (2017) grouped community interaction information due to the highest perceived risk. Information about address, phone or email (contact information) is perceived higher in sensitivity than basic demographics like gender or birth date (Phelps, Nowak, and Ferrell, 2000).

This study discloses that the willingness to disclose basic personal information like gender, birth date, and clothing size was higher among respondents compared to other information types, confirming Phelps, Nowak and Ferrell’s (2000) results. The findings also build on the findings of Milne et al. (2017). Personal information corresponds to the author’s cluster of basic demographics, where they cluster gender, birth date, and shopping behavior, due to the lowest perceived disclosure risk.

(34)

This analysis provides a first hint on the effect of the different manipulations, considering the finding that scenarios with a monetary compensation yield a higher willingness to disclose social media information.

The first regression revealed that monetary compensation is an antecedent of distributive justice. These findings confirm the construct of Ashworth and Free (2006), who state that the outcome of an information exchange is evaluated. They find that with a positive evaluated outcome distributive justice increases (Ashworth and Free, 2006). Here, monetary compensation has a positive effect on distributive justice, confirming that when a (monetary) incentive is provided in exchange for information, the outcome is perceived as fairer. Moreover, this study can also empirically validate, that a transparent privacy handling increases the perceptions of procedural justice, with procedural transparency being the only significant antecedent of procedural justice.

Drawing on reactance theory and confirming findings of the negative influence of incentives, the results of hypotheses testing also produce a significant moderation effect. When respondents feel that they are manipulated into providing personal information and they are offered a monetary compensation in exchange for their data, the positive main influence of an incentive towards perceived distributive justice is weakened. The results are in line with reactance literature and go in the same direction as the findings of Lee, Lim, Kim et al. (2015), who find incentives to mitigate concerns but at the same time state an increase of concerns when an incentive is offered for sensitive information, as respondents feel manipulated into information provision.

Examining the role of the dimensions of justice towards privacy concerns, the study disclosed that only distributive justice partially mediates the relationship between a monetary incentive towards privacy concerns. Procedural justice does not seem to influence privacy concerns. Interestingly though, a direct relationship between monetary compensation as well as between procedure transparency towards privacy concerns exists. Here, a transparent privacy

(35)

handling shows the strongest main effect on concerns, indicating that transparency can effectively lower privacy concerns. These findings go in the same direction as those of Adjerit, Acquisti, Brandimarte et al. (2013). They examined the risk that privacy policies can be used to nudge consumers into information disclosure and indeed found that a subjectively more transparent privacy policy already increases information disclosure. The results of the study at hand also confirm prior research that found incentives in exchange for information to mitigate privacy concerns (Sheehan and Hoy, 2001). Results show that concerns are also lowered when consumers consider the outcome of an information exchange as fair, which confirms one hypothesis of Ashworth and Free (2006) empirically: distributive justice is a valid construct to understand consumers privacy concerns.

Summing up, a transparent privacy handling as well as providing a monetary incentive effectively lower privacy concerns. Also, perceived distributive justice does explain privacy concerns, while procedural justice does not play an essential role in evaluating concerns. Surprisingly, privacy concerns were not found to be a significant antecedent of willingness to disclose information. With the analysis of the relationships between concerns and the three main clusters of community interaction, contact information and basic personal information, no significant main effect could be found.

Further analyses revealed though that both dimensions of distributive and procedural justice show an influence: for community interaction and contact information the two dimensions both act as a significant predictor. For personal information, only the dimension of distributive justice works as an antecedent. While the effects of the two dimensions are nearly equal with regards to community interaction, distributive justice shows a stronger effect than procedural justice on contact information. This effect stays present when looking at personal information: here, only distributive justice significantly predicts willingness to disclose information. These findings confirm the first impression gained by comparing the means and the different scenarios. It is interesting to note, that procedural fairness, which is influenced by

(36)

a transparent privacy handling makes a difference in explaining disclosure intention for higher sensitive information typologies like community interaction and contact information. A reason for this could be the fact that consumers are quite used to providing basic personal information online so that a transparent handling does not increase willingness to disclose (as the willingness is already higher than for the other information types).

Looking at the overall relationships it becomes clear that with a monetary compensation, subjects perceive the outcome of the information exchange as fair, which in turn leads to a higher willingness to disclose information. The same is true for a transparent privacy handling: it leads to a higher fairness perception, which in turn increases subject’s willingness to disclose community interaction and contact information.

As privacy concerns do not seem to play a role in evaluating which information to release, the role of a mediator of privacy concerns derived from literature could not be confirmed. This finding is contradicting previous, earlier research especially Dinev and Hart (2006), who found privacy concerns to partially mediate the relationship between perceived risks and information disclosure (Dinev and Hart, 2006). The fact that privacy concerns do not act as a relevant driver for information providing intention can be explained by a concept that gained popularity in the privacy literature over the last years (Kokolakis, 2017): The privacy paradox states, that when people are asked they are concerned about their privacy, but when it comes to actual disclosure behavior or intention, these concerns do not play a role (Norberg, 2007). Adding on to that, Marwick et al. (2016) argue that this, in turn, is not paradoxical, but that (extensive) information disclosure is a rather pragmatic response towards today’s opaque online environment, with apathy playing a more critical role than concerns (Marwick, 2016). Although there is no direct relationship of monetary compensation towards willingness to disclose information, this pragmatic behavior is also displayed in this study. When respondents were faced with a monetary incentive, they seem to evaluate the outcome, performing a cost-benefit analysis to assess if they will provide information. The strong relationship of distributive