How to influence a customer’s likelihood to allow
private data use?
Birgit Müller
S3203859
University of Groningen
Faculty of Economics and Business
MSc Marketing Intelligence
Master Thesis
June 2017
First supervisor: dr. L. Lobschat
2
A
BSTRACT
Through online technologies it is becoming easier and easier for companies to acquire private information about customers (Wirtz and Lwin 2009). At the same time, it also gets easier to misuse this data and attack a customer’s privacy. Increased customer privacy concerns lead customers to refuse private data disclosure. Past literature described the feeling of control and rewards as useful to decrease customer’s privacy concerns (Malhotra et al. 2004; Xie et al. 2006). When customers have the feeling to be able to control for what their personal information will be used for or when customers have the feeling to be rewarded when they allow a company to use their personal data, they are more likely to disclose their data. So far, it is less researched if there are interaction effects of control and reward. It is probable that the effect of reward influences the effect of control and causes together an effect which is larger than the single effect or control. Indirect effects are sometimes hidden, which is why this paper will analyze this. Furthermore, this study investigates if a restriction of the time in which the customer data can be used affects the same relationship. Differences in the perception of near and distant future decisions were found by Liberman and Trope (1998) which gives reasons to assume that the time restriction will show positive effects on the disclosure likelihood. An online study was performed to answer the formed hypotheses and research question. As expected a direct effect of control complexity on data disclosure likelihood was found but not in a shape like in the hypothesis assumed. Moreover, no moderation effects of reward or time restriction were found. However, it is worth to mention that an insignificant manipulation check made it hard to interpret the results in a reliable way. More theoretical and managerial implications as well as limitations of this study are described in the end of this paper.
Keyword: privacy concerns, control complexity, reward, bonus, time restriction, customer
3
T
ABLE OF CONTENT
Abstract ... 2 1. Introduction ... 5 2. Literature review ... 9 2.1. Privacy concerns ... 9 2.2. Control complexity ... 10 2.3. Rewards ... 13 2.4. Time restriction... 14 2.5. Conceptual model ... 16 3. Methodology ... 17 3.1. Research Design ... 17 3.2. Data Collection ... 17 3.3. Questionnaire ... 184. Descriptive statistics and reliability analysis ... 20
4.1. Descriptive statistics ... 20
4.2. Demographics ... 20
4.3. Reliability analysis and Variable distributions ... 21
4.3.1. Control Complexity ... 21
4.3.2. Decision Difficulties ... 23
4.3.3. Perceived Benefits ... 24
4.4. Manipulation check ... 24
5. Results ... 26
5.1. Hypothesis 1: private data disclosure likelihood ... 26
5.2. Hypothesis 2: Moderation effect of reward ... 29
4
6. Discussion ... 32
6.1. Control complexity ... 32
6.2. Moderation effects ... 33
7. Conclusion ... 35
7.1. Academic and managerial contributions ... 35
7.2. Limitations ... 36
7.3. Future research... 36
References ... 37
Appendix ... 39
Appendix A: Condition High Control with Reward ... 39
Appendix B: Condition Medium Control without Reward ... 41
5
1. I
NTRODUCTION
Through online technologies it is becoming easier and easier for companies to acquire private information about customers (Wirtz and Lwin 2009). Every day, people create up to 3 exabytes (1018 bytes) of data which include important information about customers that numerous companies can make use of (Sagiroglu and Sinanc 2013). It is easy for companies to collect data of their customers via a variety of sources like online searching behavior or in-store tracking (Frow et al. 2011). At the same time, it also gets easier to misuse this data and attack a customer’s information privacy. Information privacy is defined by Stone et al. (1983, p. 460) as “the ability (i.e., capacity) of the individual to control personally (vis-a-vis other individuals, groups, organizations, etc. ) information about one's self.” This information is likely to be misused by profit-driven companies with weak privacy protection (Shy and Stenbacka 2015). Selling this data without customers’ permission to third parties has become a common business (Turow et al. 2005). However, using or selling private information without a customer knowing is an infringement of privacy and can lead to loss of trust and the customer churning to a competitor (Wirtz et al. 2007). Not only customers, also privacy advocates and legislators remark privacy concerns, especially regarding privacy guarantees on the internet. Missing legal regulations for privacy guarantees on the internet increase these concerns further (Desai et al. 2003). Although, such regulations exist in the offline world, this does not serve as a reason for customers to have less concerns (European-Commission 2017).
This results in a situation where many companies have to face several problems. They know that they could satisfy their customers more efficient and use marketing efforts more effective when using customers’ private data but customers do not like to provide personal information to companies or when they do, they do not allow companies to use this data (Wirtz et al. 2007; Wirtz and Lwin 2009). So it is essential for these companies to find ways that effectively reduce the customer’s privacy concerns when asking for private data and the agreement to use it.
6 control what kind of data a company uses for what purpose privacy concerns can decrease. This feeling can be aroused on one hand by giving customers the opportunity to make their own informed choices and on the other by giving them a clear opportunity to remove their agreement and opt-out (Wirtz and Lwin 2009).
However, a company’s management has to pay attention that they do not overwhelm their customers with too much control options. Iyengar and Lepper (2000) showed with three different studies that too much choices decrease customer’s decision motivation and performance. An extensive number of choices lead to decreased intrinsic motivation and result in less qualitative outcomes [Murayama et al. (2010, p. 20911): Intrinsic motivation is the “motivation to voluntarily engage in a task for the inherent pleasure and satisfaction derived from the task itself”]. This research provides reason to assume that the positive effect of the feeling of control on the likelihood to provide a company private data is not linear, namely we expect more data to be provided when customers have the feeling to control the data use compared to situations when they cannot control it. On the other hand, further research revealed negative effects of control when too much choice is provided (Iyengar and Lepper 2000). Therefore, it is assumed that the effect of control complexity on the likelihood of private information disclosure shows an inverted U shape. In this paper control complexity is defined by the amount customers perceive to control for what purposes their personal data is used. This will be reached by presenting the information about data control in different ways. In the following, two managerially relevant ways will be described with which companies can influence the negative effect of this inverted U shape. These ways are based on literature which provided evidence for different opportunities to decrease privacy concerns.
7 gift voucher significantly increased a customer’s willingness to provide personally identifiable data.
An additional positive effect of rewards is that they increase intrinsic motivation to voluntary engage in a task (Murayama et al. 2010). These two positive effects are assumed to be able to weaken the negative effect of data overload described in the paragraph above. The perception of having too much control decreases decision motivation and performance and can lead to a decreased likelihood of private data sharing. Rewards, on the other hand, increase intrinsic motivation to complete a task and raise the willingness of data disclosure. Finally, a restriction of the period in which the private data of a customer can be used, can have an impact on the relationship between control complexity and data disclosure likelihood. When customers know that their data will be used only for a restricted period and after this period it will not be used anymore, the customers’ feeling of being in control over the data use can increase. However, customers know that the company will come back to them when the restricted period expires to ask for their permission to use the data for another period of time. Prior research showed that the anticipation of future consumption can have immediate effects on a present decision (Loewenstein 1987). It is assumed that a future event like the comeback of the company can have similar effects. Researchers found that individuals feel more accountable for the near future and for the distant future they are more optimistic (Gilovich et al. 1993; Liberman and Trope 1998). Furthermore, Sevilla et al. (2016) showed that thinking ahead to the future consumption generates positive anticipatory thoughts which will be incorporated into the customers’ current experience. It is assumed that a future event will have the same effects. In sum, it is anticipated that a time restriction of private data use will strengthen the positive effect of increasing control complexity on the one hand and that it will weaken the negative effect of too much control complexity on the other hand.
8 influences the likelihood that customers allow a company to use their private data for further purposes. The control creating variable will give customers the opportunity to decide how their private information can be used. One sub-objective of the study is to estimate the moderating effect of rewards on the effectiveness of the feeling of control on the information disclosure likelihood. A second sub-objective is to estimate the moderating effect of time restriction on the relationship between control complexity and likelihood to provide private information. This leads to the following research question:
How does the feeling of data control influence the likelihood that a customer allows private data use?
9
2. L
ITERATURE REVIEW
This chapter will give a detailed overview of the existing literature related to the research question stated in the introduction. First, an overview of customer concerns and its reasons will be given. Followed by a discussion about different ways how these concerns can be reduced. Since data misuse is one large driver of privacy concerns, providing the customer with the feeling to control how their data will be used is one opportunity to decrease privacy concerns. Furthermore, the effects of two moderators on this relationship will be stated. The first one is financial rewards which give customers the feeling to profit from disclosing private information. Secondly, will be shown how a time restriction of the period in which the company is allowed to use the data can influence the relationship between control complexity and the information disclosure likelihood. Each theoretical part will result in a hypothesis which will help to answer the research question.
2.1.
P
RIVACY CONCERNS10 disclose their real information but do not allow the company to use this information (Wirtz and Lwin 2009). For companies this situation can be a performance hurdle. They know that they could satisfy their customers more efficient and use marketing efforts more effective when using customers’ private data. So it is essential for companies to show their customers the advantages of data sharing and find a balance between customer privacy and their quests for private data and the agreement to use it (Wirtz et al. 2007).
Past research revealed different opportunities a company could use to decrease customer privacy concerns and increase private information disclosure (Awad and Krishnan 2006; Norberg et al. 2007). Some of them are: giving the customer the feeling to control the data use, offering them a reward in return and restricting the period of time within a company can use the data. Following these options will be discussed in detail.
2.2.
C
ONTROL COMPLEXITY12 satisfactory, rather than optimal”. Barry Schwartz (2004) described this phenomenon as the “Paradox of Choice”. Complex decisions force individuals “to invest time, energy, and no small amount of self-doubt, anxiety and dread” (Berry Schwartz 2004, p. 2). With an increasing number of choices the negative feelings can escalate until the customer becomes overloaded. Too much choice can become a burden resulting from complex interactions among psychological processes. Extensive choice increases expectations but also regrets; it causes opportunity costs, trade-off aversions and self-blame. Some of the main problems in a complex decision situation are the “What if” questions, the regrets and the self-blame. Regrets not only appear post-decision when the customer is unhappy with the outcome. Regrets also appear in the process of the decision making (Schwartz 2004). Questions like “Will I be happy with this option?”, “What could happen when I decide for A, for B, for C, D, E, etc.?”, “What if I take one of the other options?” can occur in a customer’s head. Customers can become too exhausted and overwhelmed from these regrets. Once regrets are in someone’s head the negative feelings continue. The customers know that they could have been able to avoid the regrettable state of affairs by choosing differently and start blaming themself (Schwartz 2004). So, too much choice in a purchase decision causes a range of negative feelings which is assumed to occur in situations with information overload, too.
The researches described so far in this chapter provide reasons to assume that the positive effect of the feeling of control on the likelihood to provide a company private data is not linear. It is assumed that the linear effect will reach a turning point from which onwards the effect of feeling more control will decrease the likelihood of private data disclosure. This is supported by further research which showed negative effects when too much choice is provided (Iyengar and Lepper 2000; Schwartz 2004). Therefore, it is assumed that the effect of the feeling to be in control on the likelihood of private information disclosure shows an inverted U shape pattern. This assumption leads to the following hypothesis:
H1: Information disclosure likelihood increases with medium levels of control complexity (vs. low levels), but decreases with high levels of control complexity.
13
2.3.
R
EWARDSAnother opportunity to treat privacy concerns is the use of rewards as exchange for private information and the agreement to use this data. This topic occurs often in comparison with the “privacy paradox”. Norberg et al. (2007, p. 100) defined it as “the relationship between individuals’ intentions to disclose personal information and their actual personal information disclosure behaviors”. Moreover, Norberg et al. (2007) showed that perceived risk has significant negative influence on the intention to disclose personal data but not on the actual disclosure behavior. On the other hand, Hann et al. (2002) showed that customers consider also the benefits they would receive form the information disclosure. This means, when thinking about information disclosure, customers take the risk of data misuse into account and compare it with the benefits from disclosing. In the actual situation of data provision it seems they do not think about the risk anymore whereby the deviation between disclosure intention and behavior is caused. So it is likely that in the situation of the decision making other things like the benefit are more important. Xie et al. (2006) showed that offering rewards, like for example a 20$ gift voucher for the customer, significantly increases private data disclosure. So when customers perceive to benefit from providing personal data their intentions to disclose further deviate from their actual disclose behavior by an increased amount of data provision (Norberg et al. 2007). The Center on Global Brand Leadership (2015) tested reward offers on two different dimensions to find the most preferred kind of rewards. On the first dimension they differed between indirect and direct rewards. Indirect rewards cannot be used for a purchase from the reward provider; an example would be a voucher for a partner company. Direct rewards are offered from the reward provider directly and can be used for purchases there like a 10% discount for the next purchase. The second dimension differs between financial and experiential rewards. Financial rewards are monetary rewards like a 10$ discount on the next purchase. An experiential reward offers an experience to the customer like a voucher for a free makeover. The results show that direct financial rewards are the most preferred ones. Therefore, it was decided to use a direct financial reward in the performed study to manipulate the perception of being rewarded when providing personal information.
14 2010). These two positive effects, the increasing effect of rewards on the private data disclosure and the positive effect of monetary rewards on the intrinsic motivation to voluntary engage in a task and finish it, are assumed to have moderating effects on the inverted U shape of the relationship between the level of control and likelihood to disclose private information. Especially, it is expected that the motivating effect of rewards will be able to weaken the negative effect of data overload described in the chapter before. The feeling of too much control decreases decision motivation and performance and can lead to a decreased likelihood of private data sharing. Rewards, on the other hand, increase intrinsic motivation to complete a task and raise the willingness of data disclosure. Therefore, the following hypothesis is formulated:
H2: Rewarding will weaken the negative effect of too much control on the likelihood of data disclosure.
Another moderator who can have effects on the inverted U shape of the relationship between control complexity and data disclosure likelihood is the restriction of the period in which a company is allowed to use the customers’ private information.
2.4.
T
IME RESTRICTION15 near future decisions the feasibility counts. As desirability the valence of an action’s end state can be seen. When it comes to distant future customers imagine the situation when the decision is over. Regarding near future decisions or feasibility, the ease or difficulty of reaching the end state is important. So customers consider the steps necessary to reach the end. Since the decision process for the restricted period of data use regards the near future, customers will consider more if they are able to run through the data providing process again when the period is over, rather than if they desire to do it. Furthermore, customers feel more accountable for near future; for the distant future they are more optimistic (Gilovich et al. 1993; Liberman and Trope 1998). Liberman and Trope (1998, p. 6) stated the reason for the optimistic distant future is that individuals “fail to incorporate factors unrelated to the task, and therefore in estimating completion time the effect of such factors is undermined.” For instance, students know well that a literature research is very time consuming but still underestimate this when planning the writing schedule for a paper. To sum up, when customers are confronted with a time restriction of the period in which their personal data can be used and therefore, with the comeback of the company to get their allowance of data use again, they will see this as a near future decision. So customers will not only consider if they are able to run through the same data provision process again, they will also feel accountable to do it. Therefore it is assumed that a time restriction of private data use will strengthen the positive effect of increasing control complexity on the one hand, and that it will weaken the negative effect of too much control complexity on the other. When customers think ahead to the situation when a company come back to get an extension of the private data use period, it is assumed that they will feel feasible and accountable to do it. This will motivate the customers to overcome the turning point of the control complexity effect and avoid a decrease caused by control overload. This lead to the following hypotheses:
H3a: The positive effect of increasing control complexity from a low to a medium level will be strengthened by the time restriction.
16
2.5.
C
ONCEPTUAL MODELTo investigate how the feeling of data control influences the likelihood that customers allow a company to use their private data for further purposes the following conceptual model was created. The hypotheses are shown in the model below and their effects are expressed with an according sign.
Figure 1: Conceptual model
17
3. M
ETHODOLOGY
This chapter will provide insights into the methods used to create and conduct the study, of which the results will be used to answer the previously formulated hypotheses and the research question. First the research design will be shown, followed by a description of the data collection process. Lastly, insights in the used questionnaire will be given.
3.1.
R
ESEARCHD
ESIGNThe aim of the study was to investigate how the feeling of data control influences the likelihood that a customer allows private data usage. In order to do so, a 3x2 between-subject design was used. The 6 conditions of this full factorial design included 3 levels of the variable control complexity (low/medium/high) and 2 levels of the variable reward (reward/no reward). In the high level of control complexity the participant was given a high amount of control over the number of purposes, namely 17, for what their personal data will be used by the company (see appendix A). In the medium and low control level the participants were given a medium (number of purposes: 4) and low (number of purposes: 2) amount of control (see appendix B and C). The variable reward was split into the levels reward, in which the participant was provided with a bonus when he or she agrees to the usage of their private data, and no reward, where no bonus was mentioned. A detailed description of the questionnaire can be found in chapter 3.3.
Low Complexity Medium Complexity High Complexity
Reward 1 3 5
No Reward 2 4 6
Table 1: 3x2 between-subject matrix
3.2.
D
ATAC
OLLECTION18 the survey was sent mainly to people living in Austria and Germany, the questionnaire was provided in both German and English.
3.3.
Q
UESTIONNAIRE19 additional questions to check if the manipulation worked. In the end of the survey each participant was asked to provide some demographic information.
20
4.
D
ESCRIPTIVE STATISTICS AND RELIABILITY ANALYSIS
Following, descriptive statistics like the total number of participants and reasons for excluding some for further analyses will be given. Moreover, demographic information of participants and the reliability analysis of survey questions will be described.
4.1.
D
ESCRIPTIVE STATISTICS278 people participated in this study of which finally 208 participants could be used for the analyses. 70 people had to be excluded because they did not fill in the survey until past the manipulation check questions or they showed signs of dishonest answers. The table below shows the total numbers of participants assigned to the different conditions. As can be seen are the number of participants in the conditions evenly distributed.
Low Complexity Medium
Complexity High Complexity Total
Reward 36 30 32 98
No Reward 35 37 38 110
Total 71 67 70 208
Table 2: Participants per condition
4.2.
D
EMOGRAPHICS21 to have over 4,500€ income per month and 67 participants (32.2%) gave no answer to this question.
Before the hypotheses testing some further data preparations needed to be done which are described in the following part of this thesis.
4.3.
R
ELIABILITY ANALYSIS ANDV
ARIABLE DISTRIBUTIONSTo check whether or not the manipulation in the study worked, the participants were asked to indicate their agreement to statements regarding the manipulation on a seven point Likert scale from 1= “I strongly disagree” to 7= “I strongly agree”. Before the manipulation check, it is necessary to check for internal consistency of all statements, whether they measure the same construct or not. Another step before the hypotheses testing is to check whether all relevant continuous variables are normally distributed.
4.3.1. Control Complexity
22 Figure 2: Control of data use of total participant sample
Searching for a reason, it was tested if outliers biased the results but no such effects were found. A more detailed look at the distributions revealed as reason for this twisted distribution two completely different perceptions of people who allowed the company “My TV” to use parts or all of their private data and those who did not allow it. As can be seen in the next graph, which shows the distribution of those 113 participants who did not allow the company to use their data, the distribution is almost similar to the figure showing the total sample.
Figure 3: Control of data use of participants who did not allow their data use
On the other side, the distribution looks completely different for those 97 participants who allowed the company to use their private data. The figure shows an almost normally distributed sample as it was expected to be for the total sample.
-5 5 15 25 35 45 55 65 1 2 3 4 5 6 7 Fre q u en cy
1= “I strongly disagree” to 7= “I strongly agree”
Control of data use
n=208 -5 5 15 25 35 45 55 65 1 2 3 4 5 6 7 Fre q u en cy
1= “I strongly disagree” to 7= “I strongly agree”
Not allowed: Control of data use
23 Figure 4: Control of data use of participants who allowed their data use
To sum up, it seems like people who did not allow the company to use their personal data did not react to the manipulations and mostly indicated 1=”I strongly disagree”. In-debt interviews with some participants showed that some of them did not perceived a feeling of control, had no difficulties to make their decisions or felt rewarded. Derived from these interviews a possible explanation for this could be that the attitude not to provide personal data is fixed in the people’s heads and is almost resistant to nudges or other manipulations.
4.3.2. Decision Difficulties
Using three other statements, it was measured how difficult it was for the participants to make their decisions. The following statements were based on scales which were already tested in the past (Bearden and Netemeyer 1999). “I had the feeling to be overwhelmed by the number of decisions about the purposes my personal data will be used”, “It was hard to decide for what purposes my personal data can be used” and “Reading through the different purposes my personal data can be used was exhausting” were used for this. A Cronbach’s alpha of 0.732 (α = .732) indicated that all three statements measure the same concept. Therefore a new average variable, measuring ‘decision difficulty’ was formed out of these three statements. The variable was normal distributed with a mean of 3.80 as well as a minimum of 1.00 and a maximum of 7.00. The mean below 4.00 indicated that it was on average rather less difficult for the participants to make their decisions.
-5 5 15 25 35 45 55 65 1 2 3 4 5 6 7 Fre q u en cy
1= “I strongly disagree” to 7= “I strongly agree”
Allowed: Control of data use
24
4.3.3. Perceived Benefits
The construct ‘perceived benefit’ was measured using again already tested scales (Bearden and Netemeyer 1999). The three statements: “I benefit from allowing „My TV“ to use my personal data”, “I receive great benefit when allowing „My TV“ to use my personal data” and “I will be rewarded when I allow "My TV" to use my personal data” were used. Internal consistency of the three statements was confirmed by a Cronbach’s alpha of 0.897 (α = .897). Again a new average variable was created ranging from 1.00 to 7.00 with a mean of 2.61. The distribution of this variable showed also a large peak in the left.
4.4.
M
ANIPULATION CHECK26
5. R
ESULTS
In the following chapter the results of the hypotheses testing will be presented and answers to the hypotheses and research question will be provided. First, the relationship between the independent variable control complexity and the dependent variable data disclosure likelihood will be investigated. In two next steps it will be analyzed if reward and time restriction have moderating effects on the relationship between the independent and dependent variable.
5.1.
H
YPOTHESIS1:
PRIVATE DATA DISCLOSURE LIKELIHOODIn Hypothesis 1, it was suggested that an increasing control complexity causes an inverted U shaped pattern of the information disclosure likelihood. At the beginning of this chapter, it will be analyzed how many of the participants in each control complexity condition allowed the company to use their data. Furthermore, it will be shown how many participants of those who perceived to have high, medium or low control disclosed their data. Next, a test of differences in the disclosure likelihood between the conditions will be done. This test was first performed on the total sample and afterwards repeated for the split sample like in the analyses before. It is worth mentioning that the unsuccessful manipulation regarding control complexity makes it hard to interpret the results in a reliable way.
27 Low Complexity Medium Complexity High Complexity
Disclosure 21.1% 43.3% 72.9%
No Disclosure 78.9% 56.7% 27.1%
Total 100% 100% 100%
Table 3: Disclosure matrix per condition
A Chi2 test confirmed that participants from the high control condition disclosed significantly more often their personal data than those from the medium (X2 (1, n=137)= 12.32; p= .000) and low condition (X2 (1, n=141)= 37.89; p= .000). Furthermore, participants from the medium condition allowed the company more often to use their data than people from the low condition (X2 (1, n=138)= 7.79; p= .005).
Next the statement “I had the feeling to control for what my personal data will be used”, which was rated on a 7 point scale from 1= “I strongly disagree” to 7= “I strongly agree”, was categorized in the levels low (participants who answered 1 or 2), medium (participants who answered from 3-5) and high (participants who answered 6 or 7). 26 participants perceived to have high control. Of these 26 people 11 (42.3%) disclosed their data and 15 (57.9%) did not. 61 participants had the feeling to have medium control. 36 (59.0%) disclosed and 25 (41.0%) did not disclose their private information. Finally, 90 participants perceived to have low control. Only 26 (28.9%) participants allowed “My TV” to use their data and 64 (71.1%) did not.
Low Complexity Medium Complexity High Complexity
Disclosure 28.9% 59.0% 42.3%
No Disclosure 71.1% 41.0% 57.79%
Total 100% 100% 100%
Table 4: Disclosure matrix per perceived level of control
28 Another test for analyzing differences between the control levels on the disclosure likelihood was performed. This time the question “How likely is it that you allow the TV provider to use your personal data?” was used. The result of the total sample analysis showed no significant differences (F (2, 205)= 1.08; p= .342). However, the results of the split sample were indeed significant. As in all tests before, the sample of participants who did not allow the company to use their data was not affected (F (2, 110)= .34; p= .713). In the sample of participants that disclosed their data, differences between the three levels were found (F (2, 92)= 10.98; p= .000). With a mean of 4.53 the low control level showed the highest data disclosure likelihood. The medium condition showed a mean of 3.90 and the high condition one of 2.57. The Bonferoni Post-hoc test showed that the differences between the high and the medium as well as between the high and the low condition were significant. The same test was performed for the perceived level of control but no reliable results could be found because there were not enough participants in the high level after the split.
29
5.2.
H
YPOTHESIS2:
M
ODERATION EFFECT OF REWARDIn Hypothesis 2 it was assumed that rewarding will weaken the negative effect of too much control on the likelihood of data disclosure. Since there was no negative effect of too much control found, this effect cannot be weakened by a reward. Thus, Hypothesis 2 cannot be supported. Nevertheless, it was interesting to see whether a reward like a bonus moderates the relationship between level of control and the private data disclosure likelihood. Therefore, regression analyses were performed to estimate which variables had direct and which had indirect effects on the dependent variable. Since this hypothesis regards the effect of a change from the medium to the high level of control, all participants from the low control condition were excluded. As in several analyses before, the results were insignificant for the total data set and for the split set including participants who did not allowed the company to use the data. In a first regression analysis the direct effects of the high level of control compared to the medium level and reward compared to no reward were investigated. The results for the participants who disclosed their data showed significant values for the whole model (F (2, 77)= 8.62; p= .000) as well as for the level of control. The beta of -1.350 shows that a participant in the high control conditions is less likely to allow the company to use his/her data compared to a participant in the medium control condition. Reward showed a marginal significant positive effect, showing that the disclosure likelihood is higher when customers get offered a reward. Moreover, the model showed a R2 value of .183. This means that the independent variables explained around 18% of the variance of the dependent variables.
Variable Beta t-value p-value
(Constant) 3.636 11.413 .000
High level of Control
Baseline: medium level -1.350 -3.732 .000
Reward .688 1.938 .056
Table 5: Linear regression analysis - direct effects of control and reward
30 relationship between independent and dependent variable. Moreover, the model showed a R2 value of .011 meaning that it explains just 1.1% of the DV’s variance. This shows that this model performs worse than the model including only direct effects.
Variable Beta t-value p-value
(Constant) 3.667 9.943 .000
High level of Control
Baseline: medium level -1.400 -3.001 .004
Reward .606 1.012 .315
High*Reward .127 .171 .865
Table 6: Linear regression analysis - moderation reward
5.3.
H
YPOTHESIS3
A AND B:
M
ODERATION EFFECT TIME RESTRICTIONHypothesis 3a assumed that the positive effect of increasing control complexity from a low to a medium level will be strengthened by the time restriction. Participants who allowed the company to use their personal data were asked how long this data can be used: restricted for 1 year or unlimited. In the performed linear regression analysis, a dummy variable was used for the time restriction of 1 year. As this question was not asked to participants who did not disclose their data, the following tests were only done for the split sample including participants who allowed the company to use their private data. Moreover, all participants from the high control condition were excluded for this test, since the hypothesis regards the effect between low and medium level of control. Both regression analyses, the one including only direct effects of level of control and time restrictions (F (2, 41)= 2.17; p= .127) as well as the one including direct and indirect effects (F (3, 40)= 1.42; p= .252) were not significant. So no moderation effect of the time restriction was found. Therefore, Hypothesis 3a is not supported.
Variable Beta t-value p-value
(Constant) 4.918 9.946 .000
Medium level of Control
Baseline: low level -.325 -.568 .573
Time restriction -.961 -1.724 .092
31
Variable Beta t-value p-value
(Constant) 4.889 8.478 .000
Medium level of Control
Baseline: low level -.264 -.314 .775
Time restriction -.889 -.975 .335
Moderator Low*Time -.117 -.101 .920
Table 8: Linear regression analysis - moderation time restriction A
Hypothesis 3b assumed that the negative effect of increasing control complexity from a medium to a high level will be weakened by the time restriction. As this negative effect was not found, a support of Hypothesis 3b is not possible. Nevertheless, an analysis was performed to see if any moderation effect was given. Participants from the low control level were excluded. Looking at the direct effects a significant model was found (F (2, 77)= 6.79; p= .002). It showed a significant negative direct effect of the high control level. Time restriction had no effects and with a R2 of .150 explains the model 15% of the DV’s variance.
Variable Beta t-value p-value
(Constant) 4.115 10.169 .000
High level of Control
Baseline: medium level -1.345 -3.641 .000
Time restriction -301 -.785 .434
Table 9: Linear regression analysis - direct effects of control and time restriction B
The model including the moderation effect of time restriction was significant, too (F (3, 76)= 5.15; p= .003). A R2 of .169 was found which shows that this model performs slightly better than the previous one. However, no moderating effect was found.
Variable Beta t-value p-value
(Constant) 4.625 8.289 .000
High level of Control
Baseline: medium level -2.096 -3.097 .003
Time restriction -1.006 -1.534 .129
Moderator High*Time 1.065 1.321 .190
32
6. D
ISCUSSION
The purpose of this study was to investigate how the feeling of data control influences the likelihood that a customer allows a company to use his/her private data. For this purpose, the relationship between level of control and data disclosure likelihood, as well as possible moderation effects of reward and time restriction, were analyzed. This part of the paper will summarize the findings of the previous chapter and relate it to the theory identified in the literature section. The table below shows an overview of all hypotheses tested in the study and whether they were supported or not. It is worth mentioning that the manipulation check of control complexity was not significant. Therefore, it is hard to interpret the results in a reliable way.
H1
Information disclosure likelihood increases with medium levels of control complexity (vs. low levels), but decreases with high levels of control complexity.
Not supported
H2 Rewarding will weaken the negative effect of too much control
on the likelihood of data disclosure. Not supported
H3a The positive effect of increasing control complexity from a low
to a medium level will be strengthened by the time restriction. Not supported H3b The negative effect of increasing control complexity from a
medium to a high level will be weakened by the time restriction. Not supported Table 11: Hypotheses overview
6.1.
C
ONTROL COMPLEXITY33 customers’ decision motivation and performance. An extensive number of choices lead to decreased intrinsic motivation and result in less qualitative outcomes. Based on this information it was assumed that the positive effect of the feeling of control on the likelihood to provide a company private data is not linear. An inverted U shape of this relationship was expected. This assumption was not confirmed by the obtained results. The analysis showed, when giving customers a list with a number of options about which data will be collected and for what it will be used, more customers provided their data when they receive a long list compared to a short one. A positive linear increase of the length of the list on data disclosure likelihood was found. However, asking customers directly how likely it is that they disclose their data turns this finding around. Participants who received a long list indicated for this question significantly lower values than participants with a medium or short list. So the result from the directly asked question showed a negative linear decrease of control complexity on data disclosure likelihood. Both results did not show an inverted U shape effect so Hypothesis 1 cannot be supported.
6.2.
M
ODERATION EFFECTSIn the literature two variables, reward and time restriction, were discovered to have a possible moderating effect on the relationship between control complexity and private data disclosure likelihood.
When customers perceive to benefit from providing personal data, their intentions to disclose deviate from their actual disclose behavior by an increased amount of data provision (Norberg et al. 2007). A reward like a bonus or a gift voucher increases a customer’s willingness to provide personally identifiable data (Xie et al. 2006). It was expected that a reward can weaken the negative effect of too much control complexity, described in Hypothesis 1 by the inverted U shape effect. Since no inverted U shape was found, there was no negative effect which reward could weaken. Therefore, Hypothesis 2 was not supported. Nevertheless, it was interesting to see if a reward moderates the relationship between control complexity and data disclosure likelihood. However, the regression analysis showed an insignificant parameter for the moderator.
35
7. C
ONCLUSION
This final chapter of the thesis will provide academic and managerial contributions. The academic value of the study will be discussed and advices to managers will be given. Moreover, limitations of this study and possible future researches will be discussed.
7.1.
A
CADEMIC AND MANAGERIAL CONTRIBUTIONSThe aim of this research was to study the relationship between the feeling of control of own personal data and the likelihood to disclose this data to a company. This study tried to fill two research gaps which were discovered in the literature chapter of this paper. Past studies have researched the direct effect of control complexity or rewarding on the likelihood to disclose data but for the best of the author’s knowledge, the interaction between control and reward was untouched. This study found huge differences of the perception of control and reward between participant who are willing to provide a company with personal information and those who are not willing. These manipulations showed direct effects on the group willing to disclose data, when analyzed separately. An interaction effect on disclosure likelihood was not found in this study. Both control complexity and reward seem to have no impact on people who are not willing to provide their personal data. To sum up, these two variables are good parameters to address customers who are likely to consider sharing their data. When a company would like to turn customers who are unwilling to share data into willing customers, these variables seem to be useless.
While researching literature, no studies were found on the topic time restriction of data usage periods in combination with data disclosure likelihood. This study discovered no significant direct effects of time restriction on the likelihood to disclose personal information. Furthermore, no moderating effect was found. So this is again a variable which is less useful when addressing people who are not likely to allow a company to use their personal information.
36
7.2.
L
IMITATIONSAs mentioned earlier, the manipulation check of control complexity turned out to be insignificant. Two possible reasons for this were discovered. First, the number of items on the control complexity list was too close between the different levels. Second, the formulation of the statement, checking this manipulation was too general. Either way, this insignificant test had influenced the generalizability of the results found in this study, and may cause further insignificant results. Furthermore, the fact that only participants who disclosed their personal data were affected by the manipulations, limited the size of the useable data set and also the number of participants per condition.
7.3.
F
UTURE RESEARCH37
R
EFERENCES
Awad, Naveen Farag and Mayuram S. Krishnan (2006), "The personalization privacy
paradox: an empirical evaluation of information transparency and the willingness to be profiled online for personalization," MIS quarterly: 13-28.
Bearden, William O and Richard G Netemeyer (1999), "Handbook of marketing scales:
Multi-item measures for marketing and consumer behavior research," California, Sage Publications.
Brandimarte, Laura, Alessandro Acquisti and George Loewenstein (2013), "Misplaced
confidences privacy and the control paradox," Social Psychological and Personality Science 4(3): 340-347.
Center-on-Global-Brand-Leadership (2015), "What Is the Future of Data Sharing? Consumer
Mindsets and the Power of Brands." Columbia Business School, Retrieved 19th March 2017, from https://www8.gsb.columbia.edu/globalbrands/research/future-of-data-sharing.
Desai, Mayur S., Thomas C. Richards and Kiran J. Desai (2003), "E-commerce policies and
customer privacy," Information Management & Computer Security 11(1): 19-27.
Dhar, Ravi (1997), "Consumer preference for a no-choice option," Journal of Consumer
Research 24(2): 215-231.
European-Commission (2017), "Protection of personal data." Retrieved 24th Feb. 2017,
from http://ec.europa.eu/justice/data-protection/.
Frow, Pennie, Adrian Payne, Ian F Wilkinson and Louise Young (2011), "Customer
management and CRM: addressing the dark side," Journal of Services Marketing 25(2): 79-89.
Gilovich, Thomas, Margaret Kerr and Victoria H Medvec (1993), "Effect of temporal
perspective on subjective confidence," Journal of personality and social psychology 64(4): 552.
Hann, Il-Horn, Kai-Lung Hui, Tom Lee and I Png (2002), "Online information privacy:
Measuring the cost-benefit trade-off," ICIS 2002 proceedings: 1.
Iyengar, Sheena S. and Mark R. Lepper (2000), "When choice is demotivating: Can one
desire too much of a good thing?," Journal of personality and social psychology 79(6): 995-1006.
Keith, Mark J, Courtenay Maynes, Paul Benjamin Lowry and Jeffry Babb (2014), "Privacy
fatigue: The effect of privacy control complexity on consumer electronic information disclosure."
Liberman, Nira and Yaacov Trope (1998), "The role of feasibility and desirability
considerations in near and distant future decisions: A test of temporal construal theory," Journal of personality and social psychology 75(1): 5-18.
Loewenstein, George (1987), "Anticipation and the valuation of delayed consumption," The
38
Malhotra, Naresh K, Sung S Kim and James Agarwal (2004), "Internet users' information
privacy concerns (IUIPC): The construct, the scale, and a causal model," Information systems research 15(4): 336-355.
Murayama, Kou, Madoka Matsumoto, Keise Izuma and Kenji Matsumoto (2010), "Neural
basis of the undermining effect of monetary reward on intrinsic motivation," Proceedings of the National Academy of Sciences 107(49): 20911-20916.
Norberg, Patricia A., Daniel R. Horne and David A. Horne (2007), "The privacy paradox:
Personal information disclosure intentions versus behaviors," Journal of Consumer Affairs 41(1): 100-126.
Nowak, Glen J and Joseph Phelps (1995), "Direct marketing and the use of individual-level
consumer information: Determining how and when “privacy” matters," Journal of Direct Marketing 9(3): 46-60.
Sagiroglu, Seref and Duygu Sinanc (2013), "Big data: A review." Collaboration Technologies
and Systems (CTS), 2013 International Conference on, IEEE.
Schwartz, Barry (2004), "The paradox of choice: Why more is less," Ecco New York.
Sevilla, Julio, Jiao Zhang and Barbara E Kahn (2016), "Anticipation of Future Variety Reduces
Satiation from Current Experiences," Journal of Marketing Research 53(6): 954-968.
Shafir, Eldar and Amos Tversky (1992), "Thinking through uncertainty: Nonconsequential
reasoning and choice," Cognitive psychology 24(4): 449-474.
Shy, Oz and Rune Stenbacka (2015), "Customer privacy and competition," Journal of
Economics & Management Strategy: 539-562.
Stone, Eugene F., Hal G. Gueutal, Donald G. Gardner and Stephen McClure (1983), "A field
experiment comparing information-privacy values, beliefs, and attitudes across several types of organizations," Journal of applied psychology 68(3): 459-468.
Turow, Joseph, Lauren Feldman and Kimberly Meltzer (2005), "Open to exploitation:
America's shoppers online and offline," A Report from the Annenberg Public Policy Center of the University of Pennsylvania.
Wirtz, Jochen and May O. Lwin (2009), "Regulatory focus theory, trust, and privacy
concern," Journal of Service Research: 190-207.
Wirtz, Jochen, May O. Lwin and Jerome D. Williams (2007), "Causes and consequences of
consumer online privacy concern," International Journal of Service Industry Management 18(4): 326-348.
Xie, En, Hock-Hai Teo and Wen Wan (2006), "Volunteering personal information on the
39
