The potential role of Construal Level Theory as an explanation for the privacy paradox

(1)

Thesis – Marketing

The potential role of Construal Level Theory as

an explanation for the privacy paradox.

Supervisor: Joris Demmers

Student Name: Jana Mueller

Student Number: 1108742

(2)

Statement of Originality

This document is written by Student Jana Muller 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)

Abstract

Privacy concern increase due to the trend of online business and the usage of people’s private information for better attraction of the target audience. Nevertheless, people are willing to give away their data which results in a contradictive behavior, the so-called privacy paradox. Consequently, the need to understand the underlying mechanism of this phenomenon gains in importance. Through two experimental online surveys with 167 participants in total, the current paper explores the research question if the construal level theory could be the underlying mechanism for this contradictive behavior. The main aim is to examine the direct effect of temporal distance on the inclination of disclosure. Additionally it endeavor to explore the potential role of the construal level as a mediator between temporal distance and the likelihood of disclosure. Mediation and correlation analyses showed no significant direct effect between the temporal distance and the inclination of disclosure. In addition, findings of the second study revealed a relation between the construal level and the perception of the consequences in the privacy calculus. However, the study failed to provide evidence for a direct effect of the construal level on the likelihood of disclosure. Further, no proof for the presence of the privacy paradox was found whereas consequently the research fails to dismiss the construal level theory as an underlying reason for the privacy paradox. These findings contribute to the understanding of the perception of the consequences in the privacy calculus. This forms a basis for further research of the connection between the construal level theory and privacy paradox.

(4)

i

Index of Figures and Tables

Figure 1 Path model - Study 1 ... 16

Figure 2 Path model - Study 2 ... 23

Table 1 Descriptive Statsistics Study 1... 14

Table 2 Means, Standard Deviations, and Correlations - Study 1 ... 15

Table 3 Statistics of direct, indirect, and total effect of temporal distance on likelihood of disclosure mediated by the construal level – Study 1 ... 16

Table 4 Descriptive Statsistics Study 2... 21

Table 5 Means, Standard Deviations, and Correlations - Study 2 ... 22

Table 6 Statistics of direct, indirect, and total effect of temporal distance on likelihood of disclosure mediated by the construal level – Study 2 ... 23

Index of Appendices

Appendix A Survey instructions for pre-test and for post-rating of open question Appendix B Experimental survey – Study 1 and 2

(6)

1

1. Introduction

“If this is the age of information, then privacy is the issue of our times.” (Acquisti, Brandimarte,

Loewenstein, 2015, p.509)

The concept of privacy evolved from the right to life, which served only physical interference with life and property, to the right to be let alone, which included also for example feelings and intellect (Warren, Brandeis, 1890). Later A.F. Westin (1967) introduced it as the right of “individuals, groups,

or institutions to determine for themselves when, how, and to what extent information about them is communicated to others” (Westin, 1967, p.7). Plenty of others extended this definition to adjust it to

newest developments and made it more concrete (Fried, 1968; Hoffman, 1980; Wang, Lee, Wang, 1998; Petty, 2000). In the privacy paradox context it would be appropriate to refer to the consumer privacy which was defined by Goodwin (1991) using two dimensions of control. First, the control over unwanted intrusions into the consumer´s environment which refers in terms of intrusions to unwanted marketing requests. The second dimension consists of the control over information disclosure.

In the information age the usage of social media, dating websites and other internet based communications lead to higher revelation of private information then before the digital age (Acquisti, Brandimarte, Loewenstein, 2015). Overall numbers of face-to-face interactions with customers are declining as a result of the trend towards online business (Awad, Krishnan, 2006). Organizations rely increasingly on the ability to analyze collected information about customers to build up a good relationship with them by giving personalized offers and attract in general new business (Culnan, Armstrong, 1999). This increase in usage of private information leads consequently to an increase of consumer’s privacy concerns (Culnan, Armstrong, 1999; Peltier, Milne, Phelps, 2009). According to a Federal Trade Commission study people are “very concerned” about their information privacy online (Pitofsky, Anthony, Thompson, Swindle, Leary, 2000).

This high privacy concerns combined with the importance and demand of private information leads to the so-called privacy paradox (Norberg, Horne, Horne, 2007) which describes the discrepancy between the privacy behavior and the privacy concerns of an individual. This can occur in various forms. In previous experiments it was found that individuals who stated that they were concerned about their privacy online revealed also personal information for small rewards as an exchange (Beresford, Kübler,

(7)

2

Preibusch, 2012; Spiekermann, Grossklags, Berendt, 2001) as for example loyalty cards / programs (Sayre, Horne, 2000). This can be seen as one manifestation of the privacy paradox next to the contradicting behavior that people are concerned about their private information but at the same time reveal these information on Facebook (Acquisti, Gross, 2006) or other social media channels.

Previous work has identified several factors that contribute to the privacy paradox. For instance it is expected that individuals underestimate the extent of the behavioral tracking online and also do not conclude to which extent the data is being used (Corrigan, 2015). Other factors are the experimenter effect (Acquisti, Grossklags, 2003) and uncertainty which arises mostly from incomplete or asymmetric information (Acquisti et al. 2015).

Next to the findings of the existence of the privacy paradox (e.g. Acquisti, 2004; Acquisti, Grossklags, 2005; Beresford et al. 2012), there are also studies which bring up doubts about the privacy paradox by demonstrating the existence of a relationship between privacy concerns and privacy protection behavior (e.g. Wakefield, 2013; Boyles, Smith, Madden, 2012). Tsai, Egelman, Cranor, and Acquisti (2007) found in their experiment and survey evidence for people willing to pay a premium to protect their privacy by using online shops that offer a high level of online protection. The privacy paradox has been well researched, often with studies yielding differing results. It is therefore imperative to analyze current theories to fully understand the nature of this subject.

As a consequence, this study is taking note of the Construal Level Theory (CLT) which links distance and level of abstraction as determinant of evaluations (Trope, Liberman, Wakslak, 2007). It predicts that a person’s mindset or construal level determines their behavior and that this construal level can be also influenced by psychological distance (Kivetz, Tyler, 2007; Trope et al. 2007). Therefore psychological distance could have an influence on the perception of the consequences of the disclosure of private information and therefore the privacy behavior.

The aim of this research is to close the gap in the present literature about the possibility that CLT could be an underlying reason for the existence of the privacy paradox by answering the question:

“To what extent can CLT explain the privacy paradox in relation to the perception of the consequences of disclosing private information?”

(8)

3

perception of psychological distance and consequently the benefits and costs of releasing personal data, therefore gather more data from the consumers. More precisely, in which way managers can formulate the questions for releasing private data online during the shopping or register process to convince more customers to disclose private information by shifting their focus to the benefits and away from the potential costs.

In two online experiments, participants will be confronted with day-to-day decisions regarding the disclosure of their private information. Each study will contain a different type of manipulation regarding the perception of psychological distance which helps to test the hypotheses which has been developed.

The remainder of this article is organized as follows, firstly a literature review will introduce the basic assumptions about the CLT and the potential link between CLT and the privacy paradox. As a result are hypotheses developed which are tested further on. Subsequently the methodology, the results and a brief discussion for each study is presented. The general discussion brings the findings of the two studies together and discuss the overall findings in a bigger extent, presents practical and theoretical implications, directions for further research, and have a closer look at the limitations which arise in this research. Following this, the final conclusions are drawn and the added value of this research is in sum presented.

2. Literature Review

This chapter presents a review of the existing literature concerning the key concepts in this thesis. This overview delivers further insights in the potential explanations for the privacy paradox which has been found in previous research. Additionally basic assumptions about the CLT are reviewed and the potential link between CLT and the privacy paradox is introduced.

2.1. Conceivable Explanations of the Privacy Paradox

According to the traditional view of economists (Becker, Murphy, 1988; Edwards, 1954; 1961; Posner, 1978; Stigler 1980), humans behave in a rational way when it comes to the decision process. Thus, people weigh pros and cons, take into account expected payoffs and costs and decide on a rational basis for the side which prevails. In terms of privacy, this is often called the privacy calculus (Laufer, Wolfe,

(9)

4

1977; Milne, Gordon, 1993; Culnan, Armstrong, 1999). Nevertheless, researchers found evidence for humans not always acting rational and that the decision-making process is often prone with all different kinds of biases which lead to an irrational choice in the process. This is the case when it comes to the privacy paradox where people indicate that they are concerned about their privacy but also reveal data very easily. The privacy paradox has already been investigated by many previous researchers which found many underlying mechanism for the contradictive behavior in the privacy paradox phenomenon. Next to others the privacy calculus (e.g. Culnan, Armstrong, 1999) was named as also the present bias and the immediate gratification bias (e.g. Acquisti, 2004).

The privacy calculus evolved out of the statement “calculus of behavior” which was introduced by Laufer and Wolfe (1977). Afterwards many researchers established term “privacy calculus” in their studies to refer to the rational process of weighting potential costs and benefits in the disclosure decision-making process (e.g. Culnan, Armstrong, 1999; Dinev, Hart, 2006; Kehr, Wentzel, Kowatsch, Fleisch, 2015). They tried to find other variables which influence the perception of the pros in this trade-off situation so they outweigh the cons and the individual would act rational by disclosing information despite the privacy concerns.

Kehr et al. (2015) stated in their research that the likelihood of disclosure can be composed by looking at the pre-existing emotional state (e.g. mood), the pre-existing attitudes (e.g. general trust in institution, privacy concerns), and the impact of these on the situational privacy calculus (e.g. perceived benefits and risks). In their research participants got set into a specific mood by using a video with sad or funny content. After this they got introduced to an app which is in need of private information to work properly. Further they got told that the data will be shared with other parties. Afterwards they were asked to disclose private information and give their opinion about general privacy concerns and institutional trust. The findings of the research suggest that the impact of pre-existing attitudes on the likelihood of disclosing private information is mediated by situational variables and therefore privacy considerations in a specific situation can overrule superordinate variables of the pre-existing attitude. Further they found that the pre-existing emotional state has an impact of the perception of the risks. This indicates that a more positive mood led to lower risk-perception than a negative mood (Kehr et al. 2015). In sum this research tried to explain the contradictive behavior of the privacy paradox by presenting

(10)

5

evidence for a situational privacy calculus which is influences by pre-existing attitudes and emotional states.

Other researchers as Acquisti (2004) tried to explain the privacy paradox by transferring the immediate gratification bias or also called the present bias to the privacy paradox. Acquisti (2004) and Acquisti and Grossklags (2005) stated for example that people weight indeed the consequences against each other but they also said that this does not happen rationally. Therefore they say that the privacy calculus is occurring but the individual is biased and therefore not acting rational. Consequently, they disagree in some extent with the privacy calculus and a rational individual of Culnan and Armstrong (1999), Kehr et al. (2015), and many others.

It is suggested that people have a systematic bias to overrate the present over the future and therefore go for immediate instead of delayed gratification (Acquisti, 2004; Acquisti and Grossklags, 2005). Many studies (e.g. Jackson, Yariv, 2014) has proven that most of the people are time inconsistent in terms of the present bias. With the assumption that the present bias is an underlying mechanism of the privacy paradox, researchers as Cheng, Shein and Chiou (2012) investigated how this present bias could be removed from the individuals mind.

Cheng et al. (2012) showed in their research that this immediate gratification can be overcome by putting the individual in a future-oriented mindset. In their research participants got split into two groups whereas one group was primed with a prospect concept. It was assumed that this would lead to overcome the immediate gratification. Subsequently, the future orientation got measured by a questionnaire. The participants in both groups had to take 42 binary choices between a certain amount of money received earlier and a larger amount received one week later to measure their delay discounting tendencies (Cheng et al. 2012). The findings suggested that the prospect concept introduces a future-related behavioral tendency and therefore weaken the effect of immediate gratification and the present bias.

Since these potential explanations can be applied to the privacy paradox but are not explaining the phenomenon under every circumstances, it is necessary to look further for an underlying mechanism of the privacy paradox.

(11)

6

2.2. Construal Level Theory

CLT links distance and abstraction with each other by introducing psychological distance as a determinant of representation, prediction, evaluation and behavior (Trope et al., 2007). Trope et al. (2007) suggest that concrete and low-level construals are unstructured, contextualized representations which include subordinate features and are used to evaluate close events. Abstract and high-level construals are in contrast schematic, decontextualized representations which include a few superordinate core features. These construals extract the basic facts for the evaluation of distant events.

Based on current literature (Trope and Liberman 2010; Trope et al. 2007; Williams, Stein, and Galguera 2014) four dimensions of psychological distance are assumed: temporal (time), spatial (space), social distance and hypothetically (probability). Psychological distance therefore implies that the further away an event or object is in terms of time or space, the more abstract an individual thinks of it (Fujita, Henderson, Eng, Trope, Liberman 2006; Liberman, Sagristano, Trope, 2002; Trope et al. 2007). Hence, if an event or object is probable, it feels psychological closer than an event that is unlikely to happen (Liberman, Förster, 2009; Wakslak, Trope, Liberman, Alony 2006). The same applies for the social distance, the closer a person is to the individual the more concrete the individual thinks about this person and vice versa (Liviatan, Trope, Liberman 2008).

According to CLT, abstract, high-level construals or mindsets are activated by events or activities which are distant and further in the future. This is congruent with the assumption that when the focus of an individual is shifting more into the future and holding therefore an abstract mindset, the preferences for delayed higher utility should increase as well (Trope, Liberman, 2000; Cheng et al. 2012). It is assumed that the relationship between distance and construal is bi-directional which indicates that the distance influences the construal but also that a manipulation in the construal level can influence the perception of distance (Trope et al. 2007).

The evaluation of a situation (e.g. the situation of disclosing private information) is based on the perceived distance and therefore on a high-level or low-level of construals. These can occur in different manifestations namely; primary vs. secondary features, feasibility vs. desirability, and pros vs. cons. Since previous research found that the privacy calculus or the trade-off between the consequences in general plays a potential role in the decision process, this research takes a closer look at the pros and

(12)

7 cons and their role according to the CLT.

At first cons are subordinated to pros, implicating that cons can only play a role if pros are present. If pros are not in the picture, the decision get not even considered and therefore thinking further about the cons of the situation gets redundant (Trope et al. 2007). This can be extended by the theory that low level construals represent the means of e.g. an action (How is it done?) whereas the high level represents the reason behind it (Why is it done?). The result of this is that pros are more abstract and cons are more concrete, since the pro is the goal of the action and the con is the side effect that occur when the action is performed (Trope et al. 2007).

Eyal, Liberman, Trope, and Walther (2004) have proven in their research that in evaluations of actions pros get more and cons less salient as psychological distance increases and consequently get cons more salient and pros less salient if distance decreases.

2.3. Construal Level Theory and the Privacy Paradox

Eyal et al. (2004) have already transferred the CLT to the decision-making process and the evaluation of actions where individuals perceive the pros and cons of an action regarding the temporal distance of the activity. Their findings suggested that pros are becoming more and cons becoming less salient with an increase in distance. Further this implies that when distance decreases, cons are more salient and pros are less salient. This is due to the finding that pros are a high-level of construal and cons are a low-level construal and that individuals rely for evaluations and behavioral intentions for distant events primarily on high-level construals (Trope et al. 2007).

In the privacy calculus, the event of receiving the benefit is perceived as immediate and the event of the disclosure of private information as more in the future. This perception can be based on findings of Trope et al. (2007, p.2) that “an event is in some manner psychologically distant whenever

it is not part of one´s direct experience,…”. Hence, the individual does not experience the impact of the

revelation of the private data at the same time as the benefit of the action which leads to a more distant perception. Therefore it can be said that the pros in the trade-off situation of the privacy calculus are perceived more immediate while the cons in the privacy calculus are perceived more distant.

It is assumed by previous researchers that this is the reason for the privacy paradox, since people go for immediate rewards rather than distant bigger rewards (immediate gratification). This refers also

(13)

8

to the so-called present bias whereas individuals tend to over-value immediate rewards at the expense of the long-term intentions (Jackson, Yariv, 2014).

This research assumes that the reason could be in the CLT as an underlying mechanism. Thus, cons are more distant which leads according to CLT to less salient cons. Hence, in the privacy calculus the relative weight of the pros increase. Therefore if people weigh the pros and cons in the privacy calculus they perceive the pro side higher than the con side and go for disclosure.

Out of this the following hypotheses can be developed:

H1a: The level of abstraction of consequences of self-disclosure will increase with psychological distance of the negative consequences taking effect.

H1b: An increase of abstraction of the consequences will result in less salient cons which directly increases the pros in their relative weight in the privacy calculus.

H1c: The less salient the cons of self-disclosure, the higher the inclination of self-disclosure.

As stated by Peltier et al. (2009), privacy is an abstract concept. Since it is assumed that individuals do not think about their privacy in a situation of disclosing private information, the assumption can be made that the construal level of the individual is more concrete. This can be ascribed to the immediate receipt of the benefits. In the disclosure decision it is assumed that the negative consequences are displayed more distant due to the uncertainty and the missing direct experience of the impact of the negative consequence. Thus, the negative consequences become less salient and the individuals weigh the positive consequences higher in the privacy calculus. Consequently, individuals think more on a concrete construal level.

This suggests that if the pros are phrased in a concrete way and the cons are presented using abstract wording, the people perceive the abstract cons as more distant and the pros more immediate whereby cons will get more salient and the indirect weight of the pros increases in the privacy calculus.

Three hypotheses will be tested to investigate the relationship between the construal level of the wording of the consequences and the perception of the meaningfulness of the consequences. Consequently this

(14)

9

study examines the impact that this relation has on the inclination of disclosure, testing the following hypotheses:

H2a: The negative consequences will become less salient as the level of abstraction of the negative consequences increases.

H2b: The inclination of disclosure will increase with an increase in abstraction of the negative consequences.

H2c: People focus more on the concrete aspects in the privacy calculus due to their concrete construal level.

3. Study 1

In this section the first study is introduced and the method and design is presented. Further findings and a brief discussion of the results is evaluated.

3.1. Methodology

Data is collected in order to investigate the effect of temporal distance and consequently the construal level of the perception of the consequences in the process of disclosing private information. In the first part of this chapter, the sample of the study is described. Further in the second part the research design and the experimental survey is outlined and the measures and variables are determined.

3.1.1. Sample

The population of this study is people who have Internet access. At first people out of the circle of acquaintances are recruited by sending emails and using other channels as for example social networking sites. Therefore, non-users of the Internet are excluded from this study. Based on the circumstances that this study investigates the privacy paradox in the online environment, it is appropriate to create a sample out of Internet users. Internet users in 2015 totaled approximately 3.17 billion people, accounting for 43% of the world’s population (The Statistics Portal, 2016a). Approximately 53% of the Internet users are between 15 and 34 years old (The Statistics Portal, 2016b). Therefore this study attempts to reach mainly the younger Internet users within this survey in order to secure a representative sample. The

(15)

10

participants in this study were in total 89 people. The minimum sample size of one treatment was fulfilled by including more than 30 participants per treatment. This minimized the probability of spurious result according to Saunders and Lewis (2012). The sample as a whole was relatively young (M=25.4, SD=2.69) whereas the youngest participant was 21 and the oldest was 36 years. In total 58 woman and 31 men participated in this study. These were originally from 23 different countries, Germany and the Netherlands were the most common origins of the represented countries in this study.

3.1.2. Study Design

To identify whether the privacy paradox can be explained by the CLT a quantitative research through an online experiment (experimental survey) was most appropriate. The survey was created by using Qualtrics. Online experiments bring along several advantages compared to offline experiments. Benefits are for example the easy access to a demographically diverse population and the cost savings which result from redundant lab space, equipment and personnel. Further, the experiment is brought to the participant themselves, providing easy access for them. Additionally, it is important to mention the generalizability of findings to different situations and therefore the assumption of a high external validity (Reips, 2000).

An experimental survey is a combination of an online survey and an experimental design with one independent variable which is manipulated to test the hypotheses that have been developed. The independent variable in this study is the temporal distance and differs between high and low temporal distances. Resulting from this, two treatments evolved in this study. The construal level is the mediating variable.

In the experimental survey the participants were randomly assigned to the different treatments. They were not aware of the existence of other treatments in the study; this secures independence. Every participant starts with a welcome text which explains what the survey is about and how long it will take. Afterwards, the participants are randomly assigned to one out of the two treatments.

In each treatment the participants are exposed to a registration form extracted from a webpage which is manipulated in the temporal distance (independent variable). This results in two conditions, the low temporal distance condition which states that “By signing up you agree to the immediate usage of

(16)

11

“By signing up you agree to the usage of your private information for marketing purpose for a project

in 2018”. It is measured if the temporal distance has an effect on the disclosing decision. More detailed

it is investigated if the changed level of abstraction in the manipulation of the distance has an effect on the perception of the cons and subsequently on the outcome of the privacy calculus and consequently the disclosing decision.

The potential mediation variable is the construal level and it is assumed that it mediates between the temporal distance and the privacy-behavior in a way that if the distance increases, the participant’s construal level also increases in their abstraction and therefore the perception of the cons changes according to CLT to less salient cons. To measure the construal level, participants are requested to state consequences for registering on a webpage with the displayed conditions using the Facebook account. In the next step they indicate if their consequence is a pro or a con. To measure the construal level of the formulated consequences, an analysis of the wording is used. This is conducted by two independent coders who are not familiar with the study and rate the words or sentences on level of abstraction (Stöber, Tepperwien, Staak, 2000).

Finally, the participants are asked to indicate how likely they are to sign up on a website with these conditions. This is measuring the dependent variable which is namely the likelihood of disclosure. The last part of the survey is exposed to every participant regardless their treatment. It consists of demographic questions, namely gender, age, and origin. Also the privacy concerns of the participants are investigated to check if the assumption of high concerns is confirmed and additional proof for the privacy paradox can be produced.

3.1.3. Measures

The likelihood of disclosure refers to the likelihood of registering on the website with Facebook. It is measured on a 7-point Likert scale (1= very unlikely, 7= very likely).

Privacy concerns are measured by using a 7-point Likert scale (1= very concerned, 7= not very

concerned).

Next to these variables, additional variables were generated by using the open questions about the perceived consequences. Therefore the variable amount of consequences is generated using a scale from zero to 10. Also the variable number of cons and pros is created. Further the valence of the first

(17)

12

named consequences (1= first named pro, 0= first named con), the total valence of the consequences (1=

very negative, 7= very positive), the percentage of the positive consequences of the total consequences the participant named are created. For this the named consequences itself and the rating of the perception of the importance of the consequence are used. These variables fulfill as in the study of Eyal et al. (2004) the purpose to measure the salience of pros and cons and indicate if positive or negative consequences are on top of mind of the participant. Thereby it can be displayed how the weight of the pros and cons are construed in the privacy calculus and if it has an impact on the intention of disclosing information by registering on the website.

The temporal distance (1= low, 0= high) is generated to distinguish between the two treatments. An analysis of the wording is used to evaluate the participant’s construal level. A similar analysis as in the study of Stöber et al. (2000) is used with the extension of using a 7-point Likert scale (1=very concrete, 7=very abstract) instead of a 5-point Likert scale. This analysis is conducted by two independent coders who are not familiar with the study. Before they begun the rating at such they had an introduction which described the task and gave two examples to explain again the difference between abstract and concrete phrasing. To obtain one final continuous value for the participant´s construal level in this survey, the mean of the construal level of all the answers of the participant and the mean between the two different ratings is generated. The IRR (Inter-Rater Reliability) is determined using the interclass correlation coefficient in SPSS. The average measure ICC was .479 with a 95% confidence interval from .359 to .576 (F(359,359) = 1.921, p<.01) and indicated a low degree of reliability between the 360 measurements.

3.2. Results

This chapter presents the results of the first study. The collected data was analyzed with IBM´s Statistical Package for Social Science version 22. General findings are presented and the developed hypotheses are tested.

3.2.1. Descriptive Statistics / General Findings

The findings of this study confirm the assumption regarding the high privacy concerns (M=2.24, SD=.85 and M=2.3, SD=.82). However, based on the low average likelihood of disclosure (M=3, SD=2.14 and

(18)

13

M=3.3, SD=2.1) no proof for the presence of the privacy paradox phenomenon was found. This was further presented in the correlation matrix (Table 2) which displayed a positive correlation between the privacy concerns and the likelihood of disclosure with r (87) = .46, p<.01. This indicates that participants who were not likely to register on the website with their Facebook account also hold high privacy concerns.

The construal level of the participants is on average ‘somewhat concrete’ (M= 3.54, SD=.84 and M=3.36, SD=1.28). In the treatment with the low temporal distance are on average slightly more consequences named than in the treatment with the high temporal distance but this difference was not significant (F (1,87) = 1.07; p=.30). In both conditions no trend for the valence of the first named consequence is seen (M=.49, SD=.5; M=.48, SD=.5). Further a binary logistic regression was performed to ascertain the effect of temporal distance on the likelihood of naming a pro or a con first. The logistic regression model was statistically not significant, χ²(1)=.012, p=.913. The total valence of the consequences are slightly negative (M=3.19, SD=1; M=3.22, SD=1.02). This aligns with the trend of the likelihood of disclosure which tended to be ‘unlikely’ (M=3, SD=2.14; M=3.3, SD=2.1). On average participants named 1.44 (SD=1.09) pros in the low temporal distance condition and 1.34 (SD=1.05) pros in the high temporal distance condition. This difference was not significant, F(1,87) = .205, p=.652.

Concerning the cons participants generated on average 2.62 (SD=1.3) negative consequences with low temporal distance and 2.32 (SD=1.25) in the high temporal distance condition. There was no significant difference between the two conditions (F(1,87) = 1.259, p=.265).

(19)

14 Table 1 Descriptive Statsistics Study 1

The correlation matrix in table 2 shows no significant correlation between temporal distance and the likelihood of disclosure (r(87)= -.07, p=.514) or any other variable.

The correlation matrix displays a positive correlation between total valence of the consequences and the percentage of pros (r(87)=.81, p<.01) and with the valence of the first named consequence (r(87)=.46,p<.01). This shows that the overall valence of the consequences was more likely to be positive for people who also stated a pro first.

Moreover, the total valence of the consequences is positively correlated with the likelihood of disclosure (r(87)=.33, p<.01). This indicates again a rational behavior of the participants in this survey, since they are more likely to reveal private information when their overall valence of their perceived consequences is more positive. Therefore the positive consequences have more weight in the trade-off in the privacy calculus than the negative consequences.

(20)

15 Table 2 Means, Standard Deviations, and Correlations - Study 1

There is no multicollinearity (VIF values < 3) between the variables likelihood of disclosure, privacy concerns, valence of first named consequence, and total valence of the consequences. There can be a small multicollinearity assumed between the variables total valence of consequences and percentage of pros (VIF vales > 3).

3.2.2. Hypotheses Testing

In the first hypothesis (H1a) it is assumed that if the distance of the negative consequences increases, the level of abstraction will increase as well. Therefore it is expected that the temporal distance has an effect on the construal level of the participant. To test all indirect and direct effects in the model the SPSS macro of Hayes (2012) is used. In table 3 the outcome of this process macro is presented.

This analysis reveals no significant results of the indirect and direct effects in the model by paying attention to the lower and upper confidence interval. Further can be said that the model explained just 2.21 (R²=.0221) percent and was not significant (F(1,85) = .9480, p=.3916).

The correlation matrix (Table 2) already indicated that there is no relation between the temporal distance and the participant’s construal level (r(87)=.08, p=.44). Thus, the hypothesis 1a can be rejected. Further table 3 displays no significant results for the direct effect between the construal level of the participant (M) and the likelihood of disclosure (Y). Also the direct (X-Y) and indirect effect is not significant which indicates no relation between the temporal distance and the likelihood of disclosure and no mediation of the participant’s construal level.

(21)

16

Table 3 Statistics of direct, indirect, and total effect of temporal distance on likelihood of disclosure mediated by the construal level – Study 1

Figure 1 Path model - Study 1

To test the second hypothesis (H1b) the relation between the participant’s construal level and the perception of the consequences is observed. More precisely, if the valence of the first named consequence, the total valence, and the percentage of pros changed in relation to the construal level of the participant. The correlation matrix in table 2 already showed no significant relations between these variables. Due to this findings is the hypothesis H1b rejected.

The final hypothesis in the first study (H1c) suggests that with less salient cons the likelihood of disclosure will increase. The number of cons did not significantly decrease with a more abstract construal level whether of the participant or the temporal distance as a test for correlation revealed. Nevertheless, the correlation matrix revealed a positive relation with the total valence and the likelihood of disclosure (r(87)=.33, p<.01).Thus, it can be confirmed that if the total valence of the consequences gets more positive, participants indicated to be more likely to reveal private information. Consequently H1c receives support by this finding and therefore is not rejected.

(22)

17

3.3. Discussion Study 1

The findings of study 1 indicated no effect of the manipulation of the temporal distance on the likelihood of disclosure. Further the temporal distance had no effect on the level of abstraction of the consequences the participant named, the amount of consequences or valence of the consequences. Hypothesis 1c proofs rational behavior of the participant by illustrating the relation between the total valence and the inclination of disclosure. This demonstrates that in the privacy calculus the pros overweighed the cons and therefore the total valence of the named consequences became more positive. Thus, the participants were more likely to reveal their data.

4. Study 2

According to the present bias individuals tend to over-value immediate rewards at the expense of the long-term intentions (Jackson, Yariv, 2014). This study tested whether the present bias or the CLT is the underlying effect of the results from the first study. This is tested by examining the effect of the construal level of the displayed consequences in the notification. The temporal distance is kept constant so that an effect will be ascribed to the construal level and not the perception of the temporal distance.

In this section the second study is introduced and the method and results are presented. In the end, the findings are briefly discussed.

4.1. Methodology

Data is collected in order to investigate the effect of the abstraction of the formulation and consequently the construal level of the perceived consequences in the process of disclosing private information. In the first part of this chapter, the sample of the study is described. Further in the second part the research design and the experimental survey is outlined and the measures and variables are determined.

4.1.1. Sample

The sample of this study was selected in the same manner as in the first study. This resulted in a total of 78 participants. The minimum sample size of the treatments was fulfilled by including more than 30 participants per treatment (Saunders, Lewis, 2012). The sample as a whole was relatively young (M=25.69, SD=3.58) whereas the youngest participant was 21 and the oldest was 40 years. In total 57

(23)

18

woman and 21 men participated in this study. These were originally from 26 different countries, Germany and the Netherlands were the most common origins of the represented countries in this study.

4.1.2. Study Design

Equally to study 1 an online experiment (experimental survey) was most appropriate using Qualtrics. In this study participants are exposed to a cookie notification which differs in the formulation of the consequences for accepting cookies. Two treatments were created to which the participants were randomly assigned.

The notifications which are presented to the participant provide consequences for accepting cookies. These pros and cons are manipulated in their wording (concrete vs. abstract). A pre-test (N=25) was conducted to measure in advance the perceived abstraction of the wording. The rating was on a 7-point Likert scale (1=very concrete; 7=very abstract). The IRR (Inter-Rater Reliability) is determined using the interclass correlation coefficient in SPSS. The average measure ICC was .983 with a 95% confidence interval from .954 to .997 (F(5,120) = 62.733, p<.01) and indicated a high degree of reliability.

Out of this pre-test the formulation of the consequences are selected which serve to manipulate the abstraction in the cookie notification. For the abstract formulations the positive consequence was rated on average with 5.36 (SD=1.551) and the negative consequence was on average 5.32 (SD=1.6). Therefore an abstract pro is “will improve the website performance” and an abstract con is “allows the

website administrator to track the onsite behavior”. For the concrete formulations the positive

consequence was rated on average 2.32 (SD=1.435). The negative consequence had a mean of 2.48 (SD=1.759). Out of the pre-test the concrete pro “will make the website run faster and smoother and

provide you with personalized offers” and the concrete con “allows the website administrator to use the tracking of your onside behavior and share it with third parties, e.g. provide online ads with information” were extracted. The notification always consists of a combination of one concrete and one

abstract consequence.

Based on this notification participants indicate their likelihood of accepting the cookie notification. The likelihood of accepting cookies is in this study the dependent variable which is called in the results later the likelihood of disclosure. The independent variable is therefore the construal level

(24)

19 of the cookie notification.

Afterwards, participants name consequences to accept cookies by themselves and rate them as a perceived pro or con on a 7-point Likert scale (1=very large con, 7=very large pro). This is used to check if the construal level of their own consequences is influenced by the construal level of the cookie notification they have been exposed beforehand.

4.1.3. Measures

The likelihood of disclosure refers to the likelihood of accepting the displayed cookies notification. It is measured on a 7-point Likert scale (1= very unlikely, 7= very likely).

Privacy concerns are measured by using a 7-point Likert scale (1= very concerned, 7= not very

concerned).

Next to these variables, additional variables were generated by using the open questions about the perceived consequences and the rating of the importance of these consequences to the participant. Therefore the variable amount of consequences is generated using a scale from zero to 10. Also the variable number of cons and pros is created. Further the valence of the first named consequences (1= first named pro, 0= first named con), the total valence of the consequences (1= very negative, 7= very positive), the percentage of the positive consequences at the total consequences the participant named, and the construal level of the manipulation (1= abstract pro, 0= concrete pro) are created.

An analysis of the wording is used to evaluate the participant’s construal level in a similar fashion to study 1 (Stöber et al. 2000). This analysis is also conducted by two independent coders who are not familiar with the study. The IRR (Inter-Rater Reliability) is determined using the interclass correlation coefficient in SPSS. The average measure ICC was .460 with a 95% confidence interval from .314 to .575 (F(265,265) = 1.864, p<.01) and indicated a low degree of reliability between the 266 measurements.

4.2. Results

This chapter presents the results of the second study. The collected data was analyzed with IBM´s Statistical Package for Social Science version 22. General findings are presented and the developed hypotheses are tested.

(25)

20

4.2.1. Descriptive Statistics / General Findings

Study 2 confirms the assumption regarding the high privacy concerns (M=2.37, SD=.63 and M=2.4, SD=.95) as study 1 did as well. The difference of the means in the two treatments was not significant (F(1,76) = .029, p=.865). However, based on the on average low likelihood of disclosure (M=4.29, SD=2.27 and M=4.1, SD=2.24) no proof for the presence of the privacy paradox phenomenon was found. This was further presented in the correlation matrix (Table 4) which displayed a positive correlation between the privacy concerns and the likelihood of disclosure with r(76) = .36, p<.01.

The number of consequences are on average 3.24 (SD=1.97) and 3.57 (SD=1.72) consequences which showed a small tendency for the concrete pros treatment to generate more consequences. However, this difference was not significant (F(1,76) = .65, p=.423). The participant´s construal level is on average ‘somewhat concrete’ (M= 3.56, SD=.68 and M=2.62, SD=1.14) and showed for the abstract pros treatment a tendency to shift more to an abstract construal level. The difference between the treatments was significant with F(1,70) =17.032; p<.01. This observation indicates that the participants align their construal level with the treatment. In the treatment with the abstract pros and concrete cons were on average more cons named as a first consequence (M=.21, SD=41) as in the concrete pros and abstract cons condition (M=.58, SD=.5). An ANOVA showed that the differences are statistically significant (F(1,76) = 12.227; p<.01). This indicates that for the participants in the treatment with abstract pros and concrete cons the negative consequences were more prevailing than the positive consequences.

The total valence of the listed consequences in both treatments is slightly negative (M=3.37, SD=1.16 and M=3.4, SD=1.11). Furthermore, there is no significant difference between the two means (F(1,76) = .013, p=.91). In total 32% of the participants do not came up with any pros and 61.5% generate maximal two pros for accepting the cookie notification.

(26)

21 Table 4 Descriptive Statsistics Study 2

The correlation matrix in table 5 displays a positive correlation between the valence of the first named consequence and the likelihood of disclosure and further the privacy concerns with for both a Pearson correlation coefficient of r(76)=.29, p<.01. This indicates that participants who named a positive consequence first are less likely to have privacy concerns and more likely to reveal their private information. The total valence of the consequences is positively correlated (r(76)=.43,p<.01) with the likelihood of disclosure. Moreover, the overall valence of the consequences for the participants who indicated to be likely to reveal private data was more positive.

As already seen in study 1, the total valence positively correlates (r(76)=.35, p<.01) with the valence of the first named consequence which indicates that people who named a pro first, were more likely to name more/stronger pros overall and result therefore in a more positive total valence.

In addition, table 5 displays a negative correlation between the construal level (manipulation) and the valence of the first named consequence with an r(76)=-.37 and the significant value less than 0.01. This could lead to the assumption that concrete pros and abstract cons lead to a higher possibility to name a positive consequence first. Regarding the participant´s construal level a negative correlation with the construal level (manipulation) is shown with an r(76)=.44 which is significant at the level 0.01.

(27)

22

This correlation indicates that participants align with the construal level of the manipulation (e.g. when the cookie notification showed abstract pros, the participants developed a more abstract mindset).

Table 5 Means, Standard Deviations, and Correlations - Study 2

There is no multicollinearity (VIF values < 3) between the variables likelihood of disclosure, privacy concerns, valence of first named consequence, construal level (manipulation), participant’s construal level, and total valence of the consequences.

4.2.2. Hypotheses Testing

The hypothesis 2a assumes that cons will get less salient if they are abstract formulated in the notification. The correlation between the construal level and the valence of the first named consequence indicates this effect by demonstrating that in the abstract cons and concrete pros treatment participants were more likely to name a pro first. Further this difference in the treatment was significant as already mentioned above. Additionally, the valence of the first named consequence is correlated with the total valence and the percentage of pros. This indicates that if a participant named a pro at first, it was likely that the total valence was more positive overall and the percentage of the pros was higher. Moreover the correlation matrix (Table 5) displayed a significant correlation between the valence of the first consequence and the number of cons (r=-.23, p<.05) which indicates that when participants named a pro at first, they named less cons in total. Consequently, support these findings the hypothesis 1a.

H2b states that the likelihood of disclosure will increase with an increase in the abstraction of the negative consequences. This assumes a mediation of the participant’s construal level as well. By using a mediation analysis it is shown that there is no direct effect between the level of abstraction in

(28)

23

the manipulation and the likelihood of disclosure. Additionally there is no significant indirect effect which rules out a mediation by the participants construal level. There is a direct effect between the construal level of the manipulation and the construal level of the participant. This was also revealed in the correlation matrix before.

Table 6 Statistics of direct, indirect, and total effect of temporal distance on likelihood of disclosure mediated by the construal level – Study 2

Figure 2 Path model - Study 2

Due to the absence of a significant direct or indirect effect between the likelihood of disclosure and the level of abstraction (construal level of the manipulation), is the hypothesis 2c rejected.

The last hypothesis (H2c) states that the participant will focus on the concrete phrased consequences which align with their construal level. Therefore it is assumed that when the wording of pros is more concrete, participants are more likely to name a pro first, name in total more pros and their total valence is more likely to be positive. Whereas in the treatment with concrete cons, it is assumed that the participant is more focused on the negative consequences and therefore is more likely to name a con first, more cons in total and that the valence of the consequences is more negative.

As the findings earlier indicated, there is a negative relation (r=-.37, p<.01) between the valence of the first named consequence and the construal level of the treatment. Hence, the concrete pros

(29)

24

treatment led to more first named pros and the concrete cons treatment evoked that more participants named a con first. The hypothesis 2c is supported by this findings.

4.3. Discussion Study 2

The results of study 2 suggest that the formulation of the consequences in for example a cookie notification have an effect on the perception of the consequences. Furthermore it was demonstrated that with an increase in the level of abstraction for the negative consequences participants named pros first which indicate that these were more on top of mind according to Eyal et al. (2004). Nevertheless, the findings did not provide evidence for a direct effect of the level of abstraction of the negative consequences on the likelihood of disclosure. Due to the support for the hypotheses 2a and 2c it can be suggested that with abstract cons and concrete phrased pros the cons get less salient. Consequently, the pros overweight the cons in the privacy calculus and therefore people are more likely to reveal their private data. However, the study failed to provide precise findings for this assumption.

5. General Discussion

This research suggests that the appearance of the privacy paradox can be explained by construal level theory as an underlying mechanism. The studies were designed to provide an answer to the research question: To what extent can construal level theory explain the privacy paradox in relation to the

perception of the consequences of disclosing private information?

The aim of this research was two-folded: firstly, it was meant to examine if the construal level and therefore a change in the temporal distance have a direct effect on the disclose-behavior of the participants in this study. Secondly, it intended to explore whether the construal level could act as mediator of the relationship between the temporal distance or the construal level of the manipulation in the treatment respectively, and the likelihood of disclosure. By this the main research question about the potential role of the CLT as an underlying mechanism of the privacy paradox could be answered.

In doing so, the current studies tested a model of individual’s perception of pros and cons of an action in regards of the temporal distance of the activity (Eyal et al. 2004). The model suggests that cons were less salient when temporal distance increases and therefore weigh the perceived pros more in the trade-off of the privacy calculus. It is further assumed that this should lead to an increased likelihood of

(30)

25

disclosure. The privacy paradox occurs in the trade-off situation between pros and cons of disclosing private information. CLT suggests that the perception of the pros and cons can be influenced by the temporal distance of the action. Therefore, the model assumes that CLT would be an explanation for the privacy paradox, cons would be less salient in the distant future and pros would get indirectly perceived more salient in the trade-off in the decision-making process. This would be reflected by a higher likelihood of disclosure if the temporal distance has increased and vice versa. The analyses of the collected data reveals that neither the temporal distance nor the participant´s construal level play a role in influencing the likelihood of disclosure. Mediation analyses points to the fact that the construal level do not mediate the relationship between the temporal distance or the construal level of the manipulation in the treatment respectively, and the likelihood of disclosure.

Study 2 shows the correlation between the participant’s construal level and the construal level of the manipulation. Participants aligned their construal level with the manipulation. Further the means of the participant’s construal level displayed that on average the participant had a somewhat concrete construal level. Out of this it can be assumed that participants had a concrete construal level in the beginning and then aligned their construal level according to the construal level of the positive consequences. This lead to the assumption that people think about the event which occurs immediate (positive consequences) when they face a trade-off regarding data disclosure. Consequently, people think in a more concrete construal level and focus on the close events namely the pros. This suggests that the focus on the pros is stronger than the alignment with the construal level, meaning that they rather align their construal level according to the pros than shifting their focus to the cons. Nevertheless, the construal level showed an effect on the valence of the consequence which is on top of mind of the person and therefore is named first.

It is observed that people who named a con first tended to have a more positive overall valence of their consequences.

There is no additional evidence for the privacy paradox found, as participants acted in a rational way and did not follow the contradictive behavior which the privacy paradox predicts. The rational behavior receives support from the findings that a more positive overall valence of the named consequences led to a higher likelihood of disclosure. This finding has to be evaluated with caution due

(31)

26

to the fact that the present study collected the observations with a survey. Therefore behavioral intention were measured and no actual behavior. According to previous researchers intention and actual behavior can differ from each other (e.g. Acquisti, Gross, 2006; Norberg et al. 2007). Furthermore no real incentive was offered due to the hypothetical situation created in a survey and not a real website, which the participants would be interested in.

The findings have important theoretical and practical implications, which will be addressed subsequently, at the same time suggesting directions for future research and acknowledging potential limitations of this research.

The current research adds to the understanding of the consumer behavior in the online world and how the decision process and therefore the consumer’s perception of the pros and cons of disclosing private information can be explained.

5.1. Theoretical Implications and Directions for Further Research

The research contributes to the extant literature by compiling data which reflects the finding about high privacy concerns of the Federal Trade Commission (Pitofsky et al., 2000). The findings regarding the privacy paradox failed to be replicated in this research. This contradicts with several researchers and their findings concerning the privacy paradox (Acquisti, Gross, 2006; Norberg et al. 2007, Spiekermann et al. 2001). Despite findings of other researchers (e.g. Wakefield, 2013; Boyles et al. 2012; Tsai et al. 2007) who had demonstrated the absence of the privacy paradox in specific situations as well, this research has no common ground with them.

Reasons for the finding of the absence of the privacy paradox could be for once because of the experimenter effect which occurs in interviews when the individual forms the respond to comply to a norm or in regards to fulfill the expectations of the researcher or interviewer (Acquisti, Grossklags, 2003). This experimenter effect would lead to a more rational behavior. This and other potential reasons are discussed in the limitations.

Moreover it was demonstrated in this study that the temporal distance of the usage of the private information has no real impact on the disclosing behavior since there was no real difference in disclosure behavior found between the two different temporal distances.

(32)

27

Further it was found that the temporal distance had no significant influence on the construal level and therefore no mediation was observed. This contradicts with the CLT (Kivetz, Tyler, 2007; Trope et al., 2007) which assumed that the construal level get more abstract when the temporal distance increases. Further, Trope et al. (2007) stated in their research that participants named more pros than cons if the temporal distance was high. This finding does not align with this study and consequently shows proof that the CLT did not played any role in the first study and therefore had no role in the privacy disclosing decision process. However, because there was no additional proof found for the privacy paradox, it cannot be ruled out that the CLT is the underlying mechanism for the privacy paradox only because of this finding.

Peltier et al. (2009) stated that privacy is in general an abstract concept which leads to the assumption that when people think more about their privacy, their construal level will get more abstract. This study assumed that individuals do not think about their privacy in the trade-off situation of disclosing private information and consequently have a more concrete construal level. Since the average construal level of the participants were ‘somewhat concrete’ this theory finds no rejection.

In addition, the second study establishes that people align their construal level of the perception of the consequences partly with the construal level of the positive consequences in the decision process of disclosing private information. This finding is in line with the construal level theory (Trope et al., 2007). However since this research failed to prove a relationship between the construal level and the likelihood of disclosure, no transfer of the mechanism of the CLT to the privacy paradox can be made.

It was demonstrated that the valence of the first named consequence of the participants had an effect on the total valence of the consequences. Thus, in case the first consequence was positive, the overall valence tended to be more positive. Further, the total valence was an indicator for the likelihood of disclosure in this study. The number of generated pros and cons had no significant relationship with the temporal distance or the wording of the cookie notification, which is contradictive with the findings of Eyal et al. (2004). They found in their study significant results for the relationship between the number of consequences and the temporal distance. Additionally, they had in their research proof of the CLT by having the majority of the participants named pros first before cons for a distant event (Eyal et al., 2004).

(33)

28

In study 1 no tendency for the first named consequence was found and therefore no additional evidence for the presence of the CLT. However, in study 2 participants named more pros at first with abstract cons and concrete pros. Therefore the level of abstraction of the consequences had a significant effect. Hence, it can be assumed that the construal level shifted the cons more in the distance due to the abstract formulation and as a result become the cons less salient and the pros indirectly more salient in the calculus.

For further research it can be suggested to investigate more the underlying mechanism of the privacy paradox. There is still no ultimate answer for this contradictive behavior which would contribute to the targeting strategy of firms and their customers. The CLT could not be proven as a reason for the contradictive behavior of people in the online environment. However, due to some aspects as the alignment of the construal level and the more positive overall consequences with abstract cons, can the CLT not completely excluded as a potential underlying theory. Further research could therefore investigate more and in different ways the relation between the privacy paradox and the construal level theory.

5.2. Practical Implications

The findings in this research yield practical implications by initially reflecting the high privacy concerns of the people. Noteworthy is that in this studies the participants acted in a rational way and aligned their privacy concerns to their likelihood of disclosure.

The research showed that the first consequence which comes into the individuals mind is an indicator of their intentions in the disclosure decision process. This indicates that if the person had originally a pro on their mind, they are likely to hold an overall positive mindset about the decision process. In practice this means that if a privacy notification brings up a positive consequence in the mind of the individual at first, it is likely to result in an overall positive opinion.

Participants in the studies indicated their behavior according to the total valence which suggests a rational balancing of pros and cons. Therefore the findings illustrate that the privacy paradox is not always applying in the decision process of revealing private information. This could be due to different limitations of this research which will be further elaborated on in the next section.

(34)

29

5.3. Limitations

The current studies contribute to existing research; however, it also has limitations. Although the minimum sample size of 30 participants per treatment (Saunders and Lewis, 2012) was fulfilled and even exceeded in all treatments, conducting a study with a larger sample size could lead to a higher number of significant results.

Finding no proof for the presence of the privacy paradox in the studies could have different reasons. The present studies had the disadvantage of examining attitudes in an unrealistic study due to the usage of a survey. The participants had more time to think about their decision and therefore did not act out of an affect. This assumption gets some reception by observations the researcher did during the data collection period. There a small number of participants were watched while answering the survey. For the second study it was observed that sometimes the participant did not realized that the picture of the notification was not an actual cookie notification. Subsequently, the participant tried to click on accept. After they realized that it was not a real notification they got further to the question regarding the likelihood of accepting the cookies. As assumed no one selected ‘very likely’. This observation indicates that the experimenter effect could be applying due to the survey situation. The fact that this research used self-reported data on the intended behavior of the participants instead of using data on their actual behavior could also be a potential explanation for this observation.

Further no real incentive was offered due to the hypothetical situation and not a real website, which the participants would be interested in. Therefore a more authentic environment could have led to a different outcome of the study.

Another limitation is that the subjective variable participant’s construal level got generated only by two people which were sometimes very different in their judgment. A greater number of people that rated the construal level of the wording of the participants may have led to a different outcome, hence increasing the IRR to a higher degree of reliability.

The last limitation of this research is the possibility that especially the participants in the second study differed in their knowledge about the impact of cookies and therefore listed less consequences. This lack of knowledge could also have influenced the phrasing of stated consequences.

(35)

30

6. Conclusion

The present research adds to the consumer behavior and privacy literature by exploring the CLT as a potential underlying mechanism of the privacy paradox. A special focus was set on the perception of the pros and cons in the privacy disclosure decision process as well as the potential role of the construal level as a mediator in this model. The results of this research show no significant effect of the temporal distance on the construal level or on the perception of the consequences. It demonstrates an alignment of the participant’s construal level with the construal level of the positive consequences in the second study. However, the main assumption concerning the influence on the likelihood of disclosure cannot be supported. Since no proof of the presence of the privacy paradox was found, the current findings cannot dismiss the construal level theory as an underlying reason for this contradictive behavior. Consequently, the privacy paradox remains a partly unresolved phenomenon in the literature of paradoxes. All things considered, there is room for further investigation of the privacy paradox and the role of the construal level theory as a potential underlying mechanism for this contradictive behavior as known as the privacy paradox.

The potential role of Construal Level Theory as an explanation for the privacy paradox

Thesis – Marketing