Abstract
Due to an increasing number of online retailers requesting the exchange of personal information in return for services, customers have become reluctant to disclose personal information on the Internet in order to protect their privacy. Several studies have revealed the growing concerns of customers regarding their online privacy. As a consequence, online retailers are forced to take action to reduce their customers' perception of the risk to increase disclosure behaviour.
For this thesis, conjoint analysis was applied to examine which particular aspects may influence the decision of a customer to disclose personal information to a retailer on a website. The results show that customers use certain cues on the retailer website as a trade-off between the benefits and risks before disclosing personal information to the retailer. More precisely, customers generally prefer to disclose their personal information to retailers having established a good reputation, as well as retailers that provide a monetary incentive in return for information, and to retailers that request only a low level of sensitive information. Moreover, websites displaying a seal of approval and/or an elaborate privacy statement are ultimately seen as more trustworthy, hence the customer is more likely to disclose his personal details.
Subsequently, moderator analysis indicates that customers who score high on privacy concerns have a higher preference for a reputable retailer who offers a discount in exchange. Furthermore, customers who score high on privacy self-efficacy have a stronger dislike to the absence of a seal of approval. Recognition of the seal moderation, indicate that customers are more willing to disclose personal information if they recognize a certain seal of approval.
These results are deemed important for all retailers who have a web shop, since this tells them that they should display certain information in order to retrieve the customers’ information for a variety of reasons.
1. Introduction
The rapid growth of mobile commerce in Europe over the past several years has made mobile commerce increasingly important, with a share of wallet of 25 per cent of all the online spending. A mobile penetration rate of 129 per cent in Western Europe suggests the potential for mobile commerce (Redactie, 2015). Various studies have revealed the growing concerns of customers about online privacy (Nam, Song, Lee, & Park, 2006; Akhter, 2014; Chen, Chen, & Mollwo, 2015). As a result, customers are becoming more reluctant to disclose personal information on the internet in order to protect their privacy (Bansal, Zahedi, & Gefen, 2015; Nilashi, Ibrahim, Reza Mirabi, Ebrahimi, & Zare, 2015).
With an increasing number of online retailers requesting for personal information, online retailers are forced to take action to decrease the perception of risk involved with disclosing personal information to the retailer in order to increase the customers’ willingness to disclose personal information (Kim, Steinfield, & Lai, 2008). Where risks is defined as “uncertainty resulting from the potential for a negative outcome” (Norberg, Horne, & Horne, 2007, pp. 106). If the perceived risks are too high, a so-called risk-barrier exists which leads consumers not to disclose their personal information, which is a cost to the firm (Wang et al., 2004).
It has hardly been used while it is so successful in predicting the influence of such attributes (Culnan & Bies, 2003). Throughout this research, a valuable piece of information will be added to the literature since this is a new framework in this research setting (Beke, Eggers, Wieringa, et al., 2016). In this study, the privacy calculus has been applied as an underlying mechanism to assess the customers’ trade-offs between risks and benefits to disclose personal information.
To distinguish which website attributes affects the customers’ perception of risk for an online retailer and thereby influences the willingness to disclose personal information, the following research question is formulated:
To what extent do website attributes that affect the perception of risk influence the customers’ decision to disclose personal information?
The aim of this paper is to find out which attributes are important factors influencing the customers’ perception of risk with regards to the privacy calculus. More specifically, this study aims to determine which risk decreasing website attributes a retailer should use to minimize the perception of risk, which consequently may lead to an increased willingness to disclose personal information. In the current marketing research literature this specific framework in combination with such an experimental setup has not been researched before (Wang et al., 2004). Although, this is very valuable to firms there is surprisingly little research conducted when it comes to the reasons for personal information disclosure behaviour in a marketing context (Beke, Eggers, Wieringa, et al., 2016), and how managers should address the customers' perception of risk (Schumann, von Wangenheim, & Groene, 2014). This research provides new insights into the field of marketing management, as it can direct companies towards optimisation of their website so as to decrease the perception of risk, thereby increasing the customers’ willingness to disclose personal information. This paper can help managers to be more aware of certain cues that can influence the willingness to disclose personal information strongly.
2. Theoretical Framework
This chapter will be used to discuss the existing literature about the website attributes that affect the customers’ willingness to disclose personal information. After the seals of approval have been identified, the privacy statements and the incentives will be presented. Secondly, the requested information by the retailer and the reputation of the retailer will be investigated. Then the moderating effect, privacy self-efficacy and privacy concerns will be discussed. Finally the conceptual model will be shown.
2.1 Seals of Approval
Based on the above-mentioned literature it is expected that the presence of a third party seal of approval on a retailer’s website will increase customers’ willingness to disclose personal information. Because a seal of approval works as a guarantee for the customer that their personal information will not be abused by the retailer. Hence, the following hypotheses have been constructed: H1a: The customers’ willingness to disclose personal information will be higher for a reputable seal of approval compared to a non-reputable seal of approval. H1b: The customers’ willingness to disclose personal information will be higher for a non-reputable seal of approval compared to no seal of approval. Despite the idea that the presence of a seal of approval lowers the perception of risk and thereby increases the customers’ willingness to disclose personal information (Kim & Kim, 2011, Oezpolat, Gao, Jank, & Viswanathan 2013), other studies have argued that this relation is limited. For example, Kimery & McCord (2002) have argued that the effect of seals of approval is limited due to the lack of awareness of the seals. Moreover, other research indicated that many customers do not recognize the seals of approval, although they understand the general function of the seals (Moores, 2005). This suggests that the effect of seals of approval will increase if customers recognize the seals. Therefore, we assume that the recognition of a seal of approval will moderate the effect on the customers’ willingness to disclose personal information, because customers may understand the function of the seal of approval if it is recognized by them (Mcknight, Kacmar, & Choudhury, 2004). Hence, the following moderating hypotheses were composed:
H1c: Recognition of the seal Webwinkelkeur will moderate the customers’ willingness to disclose personal information compared to customers who do not recognize the seal of approval.
H1d: Recognition of the seal Thuiswinkelwaarborg will moderate the customers’ willingness to disclose personal information compared to customers who do not recognize the seal of approval.
2.2 Privacy Statement
Another attribute that may influences the customers’ willingness to disclose personal information is a privacy statement (Hoffman, Novak, & Peralta, 1999; Wang et al., 2004; Kim et al., 2008). Privacy statements "refer to a description of why consumer data are gathered, how it will be used, and how it will be stored to ensure consumer privacy” (Wang et al., 2004, pp. 55). A privacy statement can influence consumers’ privacy concern (Miyazaki & Fernandez, 2000), a concern that plays a major role in the decision-making processes of the online customer (Milne & Boza 1999; Kim & Kim 2011).
Research has shown that the presence of a privacy statement may reduce the perception of risk (Wang et al., 2004), since customers are informed about how their information is used and stored (Aguirre et al., 2015; Kim et al., 2008). Consequently, customers are more willing to disclose personal information (Brandimarte et al., 2012; Hoffman et al., 1999; Kim et al., 2008; Wang et al., 2004). Several studies in this field have suggested that transparency can be considered as a sign of trustworthiness (Andrade, Eduardo, Velitchka, & Barton, 2002; Zhao, Lu, & Gupta, 2012), and “trust plays an important role in consumers’ overcoming their perceptions of risk and insecurity in online business transactions” (Nam et al., 2006, pp. 213). This implies that the level of transparency affects the perception of risk, and eventually customers’ willingness to disclose personal information. This is supported by research of Andrade et al., (2002) and Wang et al., (2004), who state that the completeness of a privacy statement reduces the perception of risk, whereas the absence of a privacy statement can lead to withholding personal information (Hoffman et al., 1999). With firms having little understanding of how these privacy disclosures affect customer privacy perceptions (Beke, Eggers, & Verhoef, 2016), additional research in this field is crucial.
Based on the above-mentioned academic literature, it is assume that the presence of a privacy statement will lead to an increased willingness to disclose personal details, since it will decrease the customers’ perception of risk. More precisely, we assume that an elaborate privacy statement is preferred over an unelaborate privacy statement, whereas any kind of privacy statement is preferred over the absence of a privacy statement. This sequence is hypothesized, since the level of transparency will be higher for an elaborate privacy statement. Hence the following hypotheses have been created: H2a: The customers’ willingness to disclose personal information will be higher for an elaborate privacy statement compared to an unelaborate privacy disclosure. H2b: The customers’ willingness to disclose personal information will be higher for an unelaborate privacy statement compared to no privacy disclosure. 2.3 Incentives
Research has indicated that incentives can be used by retailers in exchange for personal information to encourage the customer to disclose personal information (Sheehan & Hoy, 2000). Premazzi et al., (2010) mentions that these incentives can assume several forms, like cash, vouchers, coupons, lotteries or even donations, which may have different effect on the customers’ willingness to disclose personal data. Several studies have argued that even for small incentives, most customers have an increased willingness to disclose personal information (Tsai, Egelman, Cranor, & Acquisti, 2011). The incentive is taken in the cost-benefit analysis and, and even a very small incentive in return may be enough to set off the perceived risk, which will lead to customers who disclose personal information (White, 2004). In order to investigate the different effects of incentives on the willingness to disclose, this paper will take three different kinds of incentives into consideration. Specifically, monetary incentives, a one-month free trial membership and a subscription to a free newsletter, on which will be elaborate below.
Even though there are many different incentives, Deutskens (2004) mentions that a monetary incentive is the most effective in increasing response rates in online and offline surveys, because “customers can use monetary incentives flexibly for any purpose they wish, they may be perceived as providing higher benefits than a gift of the same value but of a nature determined not by the user but by the donor” (Premazzi et al., 2010, pp. 68). Based on these findings we suggest that a voucher for discount, which is the most commensurate with a monetary incentive in an on-line environment, is the most preferred option by the customer in exchange for their personal details. Moreover, research has shown that the extent of customization of the incentives in return also influences the willingness to disclose, since a more customized offer will be of a greater personal value for the customer (White, 2004). For example, “a marketer might provide free financial advising in exchange for information related to one’s financial status or free X-rated movies in exchange for information that indicates this preference” (White, 2004, pp. 44), to increase disclosure behaviour. This suggests that, if preferred, a trial membership will increases the customers’ willingness to disclose personal information, since most memberships are to some extent customizable by the user. However, we suggest that this incentive is less preferred than a voucher, since customers seek to maximize their gains (Zhao et al., 2012).
On the other hand, incentives do not always lead to a stronger intention of the customer to disclose personal information to online retailers (Premazzi et al., 2010), since customers seek to maximize benefit in return for their information and therefore only supply their details if a net gain is expected (Zhao et al., 2012). We assume that a subscription to a free newsletter is one of the smallest perceived incentives that can be offered by the retailer in exchange for the customers’ personal information. Consequently, we assume that this is the least preferred incentive in exchange for the customers’ personal data, as customers seek to maximize benefit.
assume that customers prefer a discount to a free trial membership, whereas both will be more preferred than the subscription to a free newsletter. This sequence is hypothesized because; customers are more willing to disclose personal information if the perceived benefit of the incentive increases (White, 2004). Hence, the following hypotheses have been set up: H3a: The customers’ willingness to disclose personal information will be higher if a discount will be given, when compared to a one-month free trial membership. H3b: The customers’ willingness to disclose personal information will be higher for a one-month free trail membership, when compared to a subscription to a free newsletter.
2.4 Information Requested
Another attribute found to influence the customers’ willingness to disclose personal information is the nature of the personal information requested by the retailer (Sheehan & Hoy, 2000; Phelps, Nowak, & Ferrell, 2000; Poddar, Mosteller, & Ellen, 2009; Premazzi et al. 2010). An important factor in this role of information sharing is the sensitivity of the information requested by the retailer, which is defined as “the perceived intimacy level of the information” (Premazzi et al., 2010, pp. 69). That is to say, customers perceive the disclosure of more personal information as entailing greater personal cost and are therefore less willing to disclose this kind of sensitive information (Andrade et al., 2002). This is because the perceived risk of disclosing personal information increases as the requested information becomes more sensitive (Schrammel, Köffel, & Tscheligi, 2009). Consequently, this leads to customers who are more adept of thinking strategically before disclosing their personal details, for example by falsifying, misrepresenting, withholding, or in some cases even refusing to disclose personal information (Hoffman et al., 1999; Sheehan & Hoy, 1999).
prefer to disclose demographic information over personal identifying information (Phelps et al., 2000). “Personal identifying information includes information that can be used to identify a consumer, such as a name or an e-mail address. Demographic information by itself cannot be used to identify a consumer. It can be used in aggregate, non-identifying form for market research or in conjunction with personal identifying information to create consumer profiles” (Culnan, 2000, pp. 26). The preference for demographic information is explained by Phelps et al. (2000), who implies that aggregate data is considered as less sensitive by customers compared to personal identifying information. Consequently, customers perceive a higher risk while disclosing personal identifying information (Mothersbaugh, Foxx, Beatty, & Wang, 2012).
Based on the above information we assume that the sensitivity of the information requested by the retailer influences the customers’ willingness to disclose personal information, because customers perceive an increased level of risk as the information requested becomes more sensitive (Mothersbaugh et al., 2012). Hence, the following hypotheses have been created:
H4a: The customers’ willingness to disclose personal information will be higher for a retailer that requests a low level of sensitive personal information compared to a retailer that requests a medium level of sensitive personal information. H4b: The customers’ willingness to disclose personal information will be higher for a retailer that requests a medium level of sensitive personal information compared to a retailer that requests a high level of sensitive personal information. 2.5 Retailer Reputation
deliver valued outcomes to multiple stakeholders (Kim et al., 2004). More specifically, a reputation is an evaluation of the retailer's past performance in combination with customer behaviour and can be seen as one of the main drivers of a retailer’s trustworthiness (Andrade et al., 2002; Doney & Cannon, 1997; Premazzi et al., 2010). Eventually this trustworthiness leads to a higher willingness to disclose personal information (Schoenbachler & Gordon, 2002). According to Schoenbachler & Gordon (2002), a retailer's reputation is primarily shaped by the media and word-of-mouth, which suggests that if customers hear positive evaluations about a retailer, their attitude towards that retailer will become more positive. As a consequence, consumers perceive a lower level of risk while disclosing their personal information (Sheehan & Hoy, 2000), because customers believe to know what they can expect from a retailer with a good reputation (Nepomuceno, Laroche, & Richard, 2014). Consequently, customers infer that a retailer with a good reputation does not have any reason to ruin his reputation by misusing customer privacy (Kim et al., 2004). This increased level of trustworthiness results in a decrease in the perception of risk (Andrade et al. 2002; Nepomuceno, Laroche, & Richard 2014) and this decreased level of risk leads to an increased willingness to provide such information, particularly sensitive personal information (Mothersbaugh et al., 2012).
H5a: The customers’ willingness to disclose personal information is higher for a retailer with a good reputation, when compared to a retailer with a poor reputation.
H5b: The customers’ willingness to disclose personal information is higher for a retailer with a poor reputation, when compared to a retailer without a reputation.
2.6 Privacy Self-Efficacy
Privacy self-efficacy is defined as “the perception of one’s ability to protect one’s privacy”, deemed a characteristic that affects web site privacy-related behaviors (Rifon et al., 2005, pp. 341) and therefore influences the customers’ willingness to disclose personal information (Akhter, 2014). Higher levels of online confidence relates with lower degrees of the perception of online risk, which results in customers that are less likely to rely on informational cues to reduce the perception of risk (Miyazaki & Fernandez, 2001). However, since customers vary in their perceived abilities to protect themselves online, it can be assumed that this same effect will not hold for every customers (LaRose & Rifon, 2007).
customers will have an increased willingness to engage in disclosure behaviour (Akhter, 2014).
These findings suggest that the same principle may also hold for the attribute reputation of the retailer and the attribute privacy statement, because the online experience of the customers who are high in privacy self-efficacy made them familiar with these cues (Rifon et al., 2005). As a result, these customers have an increased willingness to accept the guarantees of a seal of approval implicit (Rifon et al., 2005). Research of White (2004) found out that incentives seems to compensate for the risk an customers perceives while disclosing personal information. These findings suggest that customers high on privacy self-efficacy place less value on the incentives, since they perceive less risk while engaging in disclosure behaviour. In line with this reasoning, it also suggest that these customers place less value on the sensitivity of the information requested, because customers high on privacy self-efficacy are confident that they can deal with the potential negative outcomes involved with the disclosure of personal information (LaRose & Rifon, 2007)
We assume that a moderating effect exists between privacy self-efficacy and the customers’ willingness to disclose personal information. More precisely, we expect that customers who are high in privacy self-efficacy place less value on the seals of approval, reputation of the retailer and privacy statement. This assumption is made, because these customers are more willing to accept the guarantees of a seal of approval implicit, since they are familiar with properties of the seals. Furthermore, we assume that customers high in self-efficacy place less value on the incentives and the sensitivity of the information requested, since these customers perceive less risk and because they are confidence that they can handle the potential risks. Hence, the following hypotheses were created: H6a: If customers are high in privacy self-efficacy it will lead to a decrease in effect of the attribute reputation of the retailer on the customers’ willingness to disclose personal information.
H6b: If customers are high in privacy self-efficacy it will lead to a decrease in the effect of the attribute information requested on the customers’ willingness to disclose personal information.
H6c: If customers are high in privacy self-efficacy it will lead to a decrease in the effect of the attribute incentives on the customers’ willingness to disclose personal information.
H6d: If customers are high in privacy self-efficacy it will lead to a decrease in the effect of the attribute seals of approval on the customers’ willingness to disclose personal information.
H6e: If customers are high in privacy self-efficacy it will lead to a decrease in the effect of the attribute privacy statements on the customers’ willingness to disclose personal information.
2.7 Privacy Concerns
risk (Youn, 2009). To enhance this process retailers can inform customers about their privacy policies before the information disclosure with certain informational cues (Milne & Boza, 1999). This implies that customers high in privacy concerns are more likely to pay attention to seals of approval and privacy statements, since these informational cues explain how their personal information will be used. Next to the seals of approval and the privacy statement, research found out that the reputation of the retailer influences the level of privacy concerns as well (Nam et al., 2006). More precisely, if consumers are dealing with a retailer with a good reputation it decreases the perception of risk, because customers expect to know how their data will be used (Nepomuceno et al., 2014). Consequently, this decreased level of risk leads to an increased willingness to provide personal information (Mothersbaugh et al., 2012).
necessarily relate to privacy-protective behaviour, which might be explained by (Tsai et al. (2011) who state that even the return of little benefits for the disclosure of personal information can decrease the customers privacy concerns.
Based on the academic literature we just discussed, we assume that a moderating effect exists between the customers' privacy concerns and their willingness to disclose personal information. We disregard the contradicting findings, because research found out that almost all people consider their personal information as important and agree on the fact that it should be protected (Kyung, 2013). We assume that customers who are high on privacy concerns will prefer a reputable retailer, a low level of sensitive information requested, a discount, a seal of approval and an elaborate privacy statement while disclosing personal information, because customers who have high in privacy concerns are searching for informational cues on websites to decrease their privacy concerns. Hence, the following hypotheses have been created:
H7a: When disclosing personal information, customers who are high in privacy concerns will have an increased preference for a seal of approval.
H7b: When disclosing personal information, customers who are high in privacy concerns will have an increased preference for an elaborate privacy statement.
H7c: When disclosing personal information, customers who are high in privacy concerns will have an increased preference for a discount.
H7d: When disclosing personal information, customers who are high in privacy concerns will have an increased preference for a low level of sensitive information requested.
2.8 Conceptual Model
3. Methodology
In this chapter the methodology of this paper discussed. Firstly, an insight in the method that has been used for the data collection will be given. Subsequently the sample and procedure will be discussed. Then, an explanation of the measurement of privacy self-efficacy will be presented, which will be followed by the experimental design. Finally, the interaction effect will be discussed.
3.1 Method
In order to test the conceptual model a Dual-Response Choice Based Conjoint analysis has been used. Conjoint analysis became the most popular method for measuring customer preferences structures (Eggers & Sattler, 2011), which can be used for any type of product, website or service (Hair, 2006). A conjoint analysis is a decomposition method that makes use of an overall evaluation of a set of attributes to measure the relative preferences for each attribute and their levels. These preferences can be statistically decomposed to obtain part-worth functions. A dual response no-choice option had used to find out if the respondent was really willing to disclose personal information in the given situation. An incentive aligned Choice based conjoint has been used, to prevent any potential bias (Eggers & Sattler, 2011).
3.2 Procedure
Preference lab, which is an online survey platform for conjoint analyses (Eggers, 2016).
3.2.1 Privacy Self-efficacy
to the respondents. Table 3 shows the attributes and the attribute levels that were used in the conjoint analyses. TABLE 3 Conjoint Attributes and Levels Attribute Levels Reputation of the
retailer • • Good reputation Poor reputation • No reputation
Information
requested • • Low level of sensitive personal information requested Medium level of sensitive personal information requested • High level of sensitive personal information requested Seals of approval • Thuiswinkelwaarborg • Webwinkelkeur • No privacy seal Privacy Disclosure • • Elaborate privacy discloser Unelaborate privacy disclosure • No privacy disclosure Incentives • A 10% discount • Free trial membership • Subscription to newsletter As can be seen in table 3, the attribute ‘reputation of the retailer’ consists of three different levels of reputation ranging from a good reputation to no reputation. The same was done for the attribute ‘information requested’, which was ranging from a high level of sensitive information requested to a low level of sensitive information requested. The attribute ‘privacy disclosure’ consists out of three levels, ranging from no privacy disclosure to an elaborate privacy disclosure. Again, the same was done for the attribute ‘Seal of approval’, which also consists of tree levels: Thuiswinkelwaarborg, Webwinkelkeur and no seal of approval. Finally, the attribute ‘Incentives’ also consists of three different levels of incentives: 10 per cent discount, free trial membership and subscription to a free newsletter.
possible a baseline was created, by setting the attribute level on a neutral level (e.g. no seal of approval).
3.3 Study Design
The conjoint analysis used in the experiment is composed out of 10 choice sets, with each three options. A total of 256 website combinations where possible in this experiment and the survey was completed by 155 respondents. Since the data was clean, all choices could be used for analysis, thus there are a total of 1550 data points. In order to motivate the respondents, they had the possibility of winning a voucher of a major Dutch online retailer.
In order to assure the validity of the study we assessed several criteria as explained in Eggers & Sattler (2011). First of all, the experiment had to be balanced, which means that each attribute levels had to appear an equal number of times during the conjoint phase. Secondly, the experiment had to be orthogonal, which means that each level combination appears an equal number of times. Furthermore, the number of levels should be balanced across attributes to avoid higher importance of attributes with more levels. Finally, none of the choice sets should be dominated; all choice sets should be equally attractive to the respondent. For it to be valid it has to take these above stated assumptions into account. The software package that had been used automatically controlled for these factors. In order to create a realistic choice set, a stimulus mobile website had been created. In line with prior considerations, this study considers mobile commerce to be a subset of electronic commerce (Ngai & Gunasekaran, 2007). Appendix 1 shows an exemplary choice set.
3.4 Analysis
Where U is the utility of all respondents of the preferred situation of information disclosure (j), which is the sum of the utilities of the different attributes (β) of the specific attribute levels (X; k=1,2…K). A statistical software package had been used to estimate the utility of the attribute levels and the moderating effects.
3.5 Moderation
4 Results
In this chapter the results of the research will be presented. Firstly, the characteristics of the sample will be presented, followed by the preference estimation of the main model, with and without interaction. Finally, the hypotheses testing will be discussed.
4.1 Sample Characteristics
According to Eggers (2016), a conjoint analysis with five attributes with each three levels and ten choice sets, will require a statistical minimum of 50 respondents in order to give statistically valuable results. The experimental group consists out of 155 respondents of which 127 passed the attention test. The respondents that did not pass the attention test were removed from the sample, which resulted in a total sample of 127 respondents. As can be seen in table 4 the ratio between males an females was nicely distributed, but it is clear that it is a young age group. Almost all the respondents have a HBO degree or higher, meaning that that sample consists mainly out of young educated people. Table 4 shows an overview of the sample characteristics. TABLE 4 Sample Characteristics Percentage Number of respondents Gender Male 58.3% 74 Female 41.7% 53 Age 19-21 17.3% 22 22-24 64.6% 82 25-27 17.3% 22 28-29 0.8% 1 Education Secondary school 6.3% 8 MBO 0.8% 1 HBO 18.9% 24 Bachelor’s degree 42.5% 54 Master’s degree 31.5% 40 4.2 Dimension Reduction
Confirmatory Factor Analysis (CFA) was conducted to find out if dimension reduction was appropriate. CFA is used to test for convergent and divergent validity, whereas the Cronbach’s alpha is assed to test for internal consistency of the scales. Since the scales are adopted from prior literature, the multi-item scales had been tested before on the appropriateness of dimension reduction with CFA. Therefore, the researcher decided not conduct CFA. In addition, since SPSS does not support for CFA, Principal Component Analyses (PCA) was conducted. The test confirmed that dimension reduction for the multi-item scales privacy concerns and privacy self-efficacy was appropriate, which allowed the new variables to be created and the analysis to be done. Appendix 2 gives an overview of the factor analyses results. 4.3 Main Effects In this section the main effects will be discussed of the conjoint analysis. First, the model will be compared to the null model. Secondly, the attribute importance will be discussed before the utilities will be presented. Finally, the main effect hypothesis will be discussed. 4.3.1 Model Comparison
Before the model could be estimated, the model specifications were set for each attribute. All the attributes are at a nominal level, so all the attributes can be assumed as part worth by default. Consequently, all the assumptions are met to estimate the model. A log likelihood ratio test compares the model without parameters, the null model, with the estimated main effects only model. An overview of the model comparison can be found in table 5. The result of the log likelihood ratio test indicates that the models significantly differ from each other, since the Chi-square of 537,82 exceeds the critical value of 124,342 (df=116, α=5%). The results showed that the main effect model predicts significantly better than the null model. Hence, we will continue with the main effect model.
TABLE 5
Main effect model compared to null model
Model LL X2 Npar df R² adj AIC(LL) Hit rare
Null model -1395,2376 33,33%
Main effect model -1126,3294
4.3.3 Conjoint Analyses Results
Table 6 shows the utility estimates, the Wald values and the p-value of the attribute levels of the main effect model. The table reveals that retailer reputation, incentives, information requested, seals of approval and privacy statement have a highly significant effect on the utilities at a 1 per cent level (p= 0,000). Although the attribute levels have a significant effect on the utilities, it does not say anything about the difference between the attribute levels. In order to test the hypotheses a Z-statistic had to be obtained to find out if the different attribute levels differ significantly from each other. To use the Z-statistic the model had to be re-estimated by making use of dummy coding, instead of effect coding, to obtain the right Z-values.
In order to test H1a and H1b the difference between Thuiswinkelwaarborg (β=0,2617), Webwinkelkeur (β=0,0909) and the absence of a seal of approval (β=-0,3526) had to be investigated. The results indicate an insignificant difference between the seal Thuiswinkelwaarborg and Webwinkelkeur (Zvalue 1,5162 < -1,96). H1b states that the seal Webwinkelkeur is preferred above the absence of a seal. The test reveals that the difference between the seal Thuiswinkelwaarborg and the absence of a seal of approval is insignificant (Z-value -1,4957 < -1.96). Hence, based on the outcome of the Z-statistic, H1a and H1b are not supported.
Secondly, the attribute privacy statement was investigated. The results indicate that an elaborate privacy statement is most preferred by the customers (β=0,3700), followed by a unelaborate privacy statement (β=0,1436), whereas the absence of a privacy statement is the least preferred option (β=-0,5137). Results of the Z-statistic indicate that the difference between an elaborate privacy statement and an unelaborate privacy statement are significant (Z-value -3,222 < -1,96). The same test was used to calculate the difference between an unelaborate privacy statement and the absence of a privacy statement, which also resulted in a significant difference (Z-value 7,9595 > 1,96). Hence, H2a and H2b are both supported.
The results in table 6 reveal that customers prefer a discount the most (β=0,3383), followed by a trial membership (β=0,0031) and a subscription to a newsletter (β=-0,3414). H3a states that customers would prefer a discount to a trial membership, which is supported by a significant difference between the two (Z-value 4,4176 > 1,96). The same method was applied to investigate the difference between the attribute levels one-month free trial membership and a subscription to a free newsletter. The test revealed that there is a significant difference between the two levels (Z-value 8,7046 > 1,96). Hence, H3a and H3b are supported.
As can be seen in table 6, customers prefer a low level of information requested (β=0,6002) the most, followed by a medium level of information requested (β=0,0016) and large level of information requested (β=-0,6019). In order to test H4a, the difference between a low level of sensitive information request and a medium level of sensitive information request had to be investigated. The results indicate that there is a significant difference between the two attribute levels (Z-value -8,5659 < -1,96). Moreover, the difference between a low level of sensitive information requested and a medium level of sensitive information requested was investigated, which again resulted in a significant difference (Z-value 6,7054 > 1,96). Hence, H4a and H4b are supported.
Finally, H5a and H5b are investigated. The results show that Bol.com (β=0,3561) is the most preferred attribute level for the reputation of the retailer, followed by Baleno (β=-0,0869) and no reputation (β=-0,2691). In order to test H5a, the difference between Bol.com and Baleno was calculated. The results indicated that the difference is significant (Z-value 5,936 > 1,96). The same was done for the attribute levels Baleno and no reputation, which also resulted in a significant outcome (Z-value 7,902 > 1,96). Hence, H5a and H5b are supported.
TABLE 6
Utility Estimations of the main Effects only Model
Attributes Utility Wald P-value
Retailer reputation No reputation -0,2691 71,5933 0,000* Bol.com 0,3561 Baleno -0,0869 Incentive Subscription to free newsletter -0,3414 75,7467 0,000* A 10% discount 0,3383 A one-month trial free membership 0,0031 Information requested High level of info requested -0,6019 220,2765 0,000* Medium level of info requested 0,0016 Low level of info requested 0,6002 Seals of approval Thuiswinkelwaarborg 0,2617 60,1243 0,000* Webwinkelkeur 0,0909 No seal of approval -0,3526 Privacy Statement Elaborate privacy statement 0,3700 113,8504 0,000* Unelaborate privacy statement 0,1436 No privacy statement -0,5137 No-choice option -0,0125 0,000* * Significant at an 1% level 4.3.2 Utility Levels
FIGURE 2 Part-worth Utility 4.3.4 Attribute Importance To find out to what extent each attribute influences the decision of the customer to disclose personal information, the range within each attribute had to be calculated. The results can be found in figure 3. The dual response no-choice option has not been take into consideration in the figure, since this does not affect the attribute importance. The results reveal that the most important attributes in the respondents’ decision to disclose personal information is the amount of information requested (29,83%), followed by the attribute privacy statement (21,93%), Incentive (16,87%), Reputation (15,51%) and privacy seal (15,24%). FIGURE 3 Attribute Importance -0,8 -0,6 -0,4 -0,2 0 0,2 0,4 0,6 0,8 Utility 15,51% 16,87% 29,83% 15,24% 21,93% 0% 5% 10% 15% 20% 25% 30% 35% Company
reputation Incnetive Information requested approval Seals of statement Privacy
Im
p
or
ta
4.4 Moderating Effects
In this section the moderating effects of the conjoint analysis will be discussed. First, the model without moderation will be compared to the model with moderation. Secondly, the attribute importance will be discussed before the utilities will be presented. Finally, the hypotheses will be tested.
4.4.1 Model Comparison
In order to test the hypotheses properly a goodness of fit comparison was conducted between the main effect model and the model with moderation to find out which model has the best fit with the data. Table 7 gives an overview of the results. The results show that the model with moderation has got a lower has a lower Log likelihood compared to the main effect model. In addition, the adjusted R-squared and the hit rates show a small improvement for the model with interaction compared to the main effect model. However if the information criteria are assessed, one can conclude that the main effect model has got a better fit with the data. Although, the data shows a better fit with the main effect model this research continues with the model with moderation, since this model gives the opportunity to test the moderating effect. TABLE 7 Fit of the Models
Model LL(0) LL(β) Npar R² adj AIC(LL) BIC(LL) Hit rate
Model with moderation -1395,2376 -1106,0141 94 0,218 2278,0282 2371,8864 59,92% Main effect model -1395,2376 -1126,3294 116 0,2024 2274,6588 2305,9448 59,13% 4.4.3 Conjoint Analyses Results Table 8 shows the estimated utilities and the significance levels of the parameters (for all values, please take a look at Appendix 3). To assess the stability of the results, first the main effect model was estimated; where in each additional step a moderator was added to conclude with the full model, model four. The results show (Appendix 4) that the data is stable, since the models do not show large differences between each other.
The full model in table 8 shows many insignificant effects therefore, a model had been created with only the significant variables (Appendix 5). To find out if the new model fits the data better the information criteria had been investigated. The results (Appendix 6) indicate that the significant-only model has the best fit with the data. Despite these results, we continue with the full model in order to be able to interpret all the results to test the hypotheses.
TABLE 8 Overview Utilities
Model one Model two Model three Model four
significant difference (p<,000). Hence, H1c is supported and H1d is not supported by the data. Next, In order to test H6a-H6e the interaction effects between privacy self-efficacy and the attribute levels had to be investigated. The hypotheses state that customers who have a high score on privacy self-efficacy do not have a preference for one of the retailers with a reputation, preference for one the seals of approval and preference for one of the privacy statements. Moreover, H6c and H6d state that customers high in privacy self-efficacy do not have a preference for a certain incentive or a certain sensitivity of the information requested. The utility estimates of the reference had to be calculated manually; therefore the significance levels of the reference levels are not reported. Unfortunately the statistical software that was used, did not account for this problem. Hence, the reference levels are considered as insignificant, since the corresponding significance levels are not reported.
Next to the interaction effect between privacy self-efficacy and the attribute levels, the same is done for the moderator privacy concerns. H7a states that customers prefer a seal of approval while disclosing personal information. The results indicate an insignificant interaction effect between privacy concerns and Thuiswinkelwaarborg (Utility=0,3280, p=0,21) and Webwinkelkeur (Utility=0,1773, p=0,57). Hence, the results indicate that H7a is not supported. To investigate H7b the interaction effect between privacy concerns and an elaborate privacy statement has been investigated. As can be seen in table 8, the interaction effect is insignificant (utility=0,3692, P=0,79). Hence, from the results can be concluded that H7b is not supported. The interaction effect between privacy concerns and discount show a partially significant effect (utility=0,2618, p=0,076). Consequently, H7c is partially supported. Next, to investigate H7d the interaction effect between the moderator and a low level information requested (utility=1,2164, p=N/A) and a medium level of information requested (utility=-0,0478, p=0,42), which resulted in an insignificant effect. Hence H7d is not supported. Finally, the interaction effect for privacy concerns and Bol.com (utility=0,2775, p=0,073) is investigated. Since the interaction effect is partly significant it can be concluded that H7e is partly supported. Customers prefer Bol.com to disclose personal information.
4.4.2 Utility Levels
FIGURE 5 Utilities of the Information Requested FIGURE 6 Utilities of the Incentives FIGURE 7 Utilities of the Seal of Approval FIGURE 8 Utilities of Privacy Statement 4.4.4 Attribute Importance In order to see which attributes are considered the most important by the customers the attributes importance of the main- and moderator effects are plotted. To do so, -0,5 -0,4 -0,3 -0,2 -0,1 0 0,1 0,2 0,3 0,4 0,5 No
brand Baleno Bol
Utility Main effect Privacy concerns Privacy self-efRicacy -1 -0,5 0 0,5 1 1,5 Request
large medium Request Request little
the range within each attribute was to be calculated. The results can be found in figure 9.
Figure 9 shows a relatively large difference in the attribute importance between the moderating variables and the main effect. The most notable differences are in the attribute information requested. Customers high on privacy concerns attach the most value on the attribute information requested (42,46%), followed by the main effect (28,4%) and customers high on privacy self-efficacy (26,99%). Additionally, customers high on privacy concerns attach the most value to the privacy statement (27,13%) compared to customers high on privacy self-efficacy (22,62%) and the main effect (20,01%). Customers high on privacy self-efficacy seem to attach the most value to the attribute company reputation (18,63%) and incentives (18,63%). On the other hand, customers high on privacy concerns (10,56%) and customers high on privacy self-efficacy (12,98%) attach less value to the seals of approval compared to the main effect (17,75%). FIGURE 9 Attribute Importance 4.5 Hypothesis testing Summary TABLE 9 Summary of Hypotheses Testing Results
Variable Hypothesis Conclusion
0% 5% 10% 15% 20% 25% 30% 35% 40% 45% Company
reputation Incentives requested Info approval Seal of statement Privacy
H1b Not supported
Recognition of Webwinkelkeur * Webwinkelkeur H1c Supported
Recognition of Thuiswinkelwaarborg * Thuiswinkelwaarborg H1d Not supported
Privacy statement H2a Supported
H2b Supported
Incentives H3a Supported
H3b Supported
Information requested H4a, Supported
H4b Supported
Retailer reputation H5a Supported
H5b Supported
Privacy self-efficacy * Seals of approvals H6a Supported
Privacy self-efficacy * Privacy statement H6b Supported
Privacy self-efficacy * Incentives H6c Supported
Privacy self-efficacy * Little sensitive information requested H6d Supported
Privacy self-efficacy * retailer reputation H6e Supported
Privacy concerns * Seals of approval H7a Not supported
Privacy concerns * Elaborate privacy statement H7b Not supported
Privacy concerns * Discount H7c Partially supported
Privacy concerns * Little sensitive information requested H7d Not supported
Privacy concerns * Bol.com H7e Partially supported
5. Discussion
This present chapter will discuss the findings of this research. Firstly, the theoretical implications of the main effect and the interaction effects will be discussed. Secondly, the managerial implications will be given. Finally, the implications of the results will be discussed, followed by the limitations of this paper as well as suggestions for further research. 5.1 Theoretical Implications 5.1.1 Main Effects This research is the first in the field to investigate a complete set of attributes on a website so as to predict the customers’ willingness to disclose personal information beforehand. The results reveal that all the hypotheses of the main effects were in line with the expectations based on previous literature. Except, for the hypotheses regarding the seals of approval, which have shown notable outcomes. These particular findings shall be discussed to a greater extent later on in this chapter.
The results show that the attribute levels Thuiswinkelwaarborg, Webwinkelkeur have a positive effect on the utility estimates, whereas the absence of a seal of approval will have a significant negative effect on the utilities. Based on prior literature, it was expected that customers would have a preference for one of the seals of approval, however the Z-statistic revealed that the differences between the attribute levels are insignificant. This indicates that customers do not have a significantly different preference for one of the privacy seals that have a positive effect on disclosure behaviour. These contradictory results may be explained by research of Kimery & McCord (2002), who state that the effect of a seal of approval is limited due to a lack of understanding and awareness of the seals. This suggests that customers do not hold a preference for a certain seal of approval, because they may not understand the difference between these seals. Consequently, the absence of a seal of approval will result in a decreased willingness to disclose personal information, and the presence of a seal of approval will increase this effect.
Moreover, in line with our hypotheses customers prefer an elaborate privacy statement as compared to an unelaborate privacy statement, and both are preferred over the absence of a privacy statement. This indicates that customers prefer to know how their data will be stored to enable their privacy. Moreover, the results suggest that an equivalent of monetary incentives (e.g. discount or a one-month free trial membership) generally constitute the most preferred incentives by the customer while disclosing personal information, whereas a subscription to a free newsletter was disliked. Customers prefer an equivalent of a monetary incentive since this pays off directly and may be seen as a fair payoff that compensates for the risks, whereas the contrary seems to be applicable for a subscription for a free newsletter.
In line with prior findings of Premazzi et al. (2010) this paper found a negative relationship between the amount of sensitive information requested by the retailer and the customers’ willingness to disclose personal information. This coincides with prior research which has stated that the potential risks of disclosing personal information increases as more sensitive information is requested by the retailer (Schrammel et al., 2009). Finally, in accordance with the hypotheses, this research found that customers prefer a retailer with a good reputation compared to a retailer with poor reputations, whereas either is preferred over an unknown retailer without a reputation. These findings are supported by Kim et al. (2004), who state that customers know what to expect from a retailer with an established reputation, consequently reducing the perceived risk while disclosing personal information. Accordingly, customers are more willing to disclose personal information (Mothersbaugh et al., 2012).
5.1.2 Privacy Self-efficacy
The results indicate that customers who score high on privacy self-efficacy perceive the absence of a privacy statement as more negative compared to customers who do not score highly on this scale. A possible explanation for these results may be given by Miyazaki & Fernandez (2001), who state that customers who are high in privacy self-efficacy have more confidence while browsing online, because they perceive less risk compared to customers who are not high in self-efficacy. Consequently, these customers are less likely to rely on informational, since they are familiar with the properties of the cues and because they are confident that they can handle the potential negative outcomes. This implies that these customers take the guarantees of informational cues implicitly, which suggests that these customers check whether certain cues are present on the webpage, whereas only the absence of these cues can negatively affect the customers’ willingness to engage in disclosure behaviour. Furthermore, this research found no moderating effect for the attribute incentives. This indicates that customers who score high on privacy self-efficacy do not have a preference for a certain incentive compared to customers who are not high in self-efficacy. An explanation for these results is given by White (2004), who suggests that an incentive seems to compensate for the risk a customer perceives while disclosing personal information. However, since customers who are high in privacy self-efficacy do not perceive a lot of risk, even the smallest incentive seems to compensate for the potential risks. In line with this reasoning, customers that score high on privacy self-efficacy also do not have a preference for the sensitivity of information requested by the retailer. It seems that the extent of sensitivity of the information requested does not matter to the customer, because customers who score high on privacy self-efficacy are confident that they can cope with the potential negative outcomes (LaRose & Rifon, 2007).
5.1.2 Privacy Concerns
disclose information to a retailer with a good reputation. On the contrary, the results show insignificant interaction effects between the moderator privacy concerns, seals of approval and privacy statements. These findings contradict the previous academic literature, which essentially argued that customers who are high on privacy concerns are likely to show risk-coping behaviour to reduce the perception of risk, while disclosing personal information (Youn, 2009). In doing so, these customers will pay more attention to informational cues, such as seals of approval and privacy statements (LaRose & Rifon, 2007). These findings are noteworthy, because they suggest that customers who scores high on privacy concerns do not pay extra attention to these informational cues to reduce their perception of risk. In addition, the sensitivity of information requested by the retailer seems to have no significant effect on the disclosure behaviour of customers who score high on privacy concerns. The findings are counterintuitive, because one would expect that these customers hold a preference for the amount of information requested, since this in turn influences the level of perceived risk. The results suggest that customers who score high on privacy concerns do not perceive the exchange of sensitive information as more negative compared to customers who score low on this scale.
5.1.2 Recognition of the Seals
Finally, the moderating effects between recognition of the seal and the corresponding attribute levels have been investigated. The results indicate that customers who recognize the seal Webwinkelkeur are more likely to disclose personal information compared to customers who do not. On the other hand customers who recognize the seal Thuiswinkelwaarborg cannot be said to be more likely to disclose personal information. These contradicting findings can be explained by research by Kim et al., (2004), which states that there is a lack of understanding and awareness for seals of approvals. This suggests that customers have a more positive perception of the seal Webwinkelkeur, indicating that this seal is perceived as a better guarantee that the customers’ privacy will not be misused compared to the seal Thuiswinkelwaarborg (Kim & Kim 2011). Hence, essentially, recognition of the seal Webwinkelkeur has a positive effect on the willingness to disclose personal information. 5.2 Managerial Implications Besides the theoretical implications, there are also several managerial implications for online retailers who would like to collect personal information for a variety of commercial reasons. One of our main findings indicate that retailers should place a seal of approval on their website, since the presence of a seal of approval will stimulate disclosure behaviour, where the seal Thuiswinkelwaarborg is typically the most preferred by customers.
With the attribute information requested by the retailer being the most important perceived attribute, managers should be attentive to the sensitivity of the information requested. This research illustrates that customers tend to prefer the smallest amount of personal information requested by the retailer. However, this can most likely be countered by a higher perceived (monetary) incentive in return, since customers increasingly perceive more risk while the retailer requests a higher amount of sensitive information. To offset these perceived risks, the customer needs a higher perceived incentive in exchange for their personal information. This research has revealed that customers generally perceive a privacy statement as the second most important factor in their trade-off to engage in disclosure behaviour. As a result, managers should carefully think about their privacy policy and specify this explicitly in a privacy statement on their website, seeing as how the presence of a privacy statement increases the customers’ willingness to disclose personal information. The extent of the privacy statement should not be a restrictive factor to managers, since an elaborate privacy statement is preferred over a concise one. Ultimately, this research has revealed that altogether customer prefers to disclose personal information to a retailer with an established reputation, because customers know what they can expect from such a retailer. However, this finding should not discourage new entrants, as this is the second to last attribute that customers take into consideration in their trade-off to disclose personal information.
5.3 Limitations and further Research Due to the scope of this present paper, several limitations can be identified in our investigation, which would require further research. First of all, the sample consists solely of students within the age range of 19 to 29 year old. The limitation of our sample size may have affected the generalizability of this experiment. We can expect that these people are generally intelligent and tech-savvy people, so we expect the results to be somewhat biased, since I expect that these people are more self-conscious online, when compared to older people. If we had included even younger people, this effect might be even stronger. On the contrary this effect may decrease if the sample consisted out of older people, since these people are generally less tech-savvy. Further research would have to be conducted to find out whether the results hold for a larger sample with a wider age range of metadata.
Another shortcoming of this research may be the number of attributes in combination with a relatively high number of choice sets in the survey. The possibility remains that the respondents were unable to pay attention the whole time, since the stimuli where quite similar to each other. As a result, respondents might have lost concentration towards the end, which may have influenced the results of this experiment. Further research is recommended in order to investigate whether the number of attributes actually influenced the results of this research. Another possible solution to measure the extent of concentration would be to estimate two models with the data, with the first models consisting of the first half and the second model consisting of the second half of the choice set. The researcher didn’t apply this method, since the exclusion of the respondents who did not pass the attention test resulted in marginal differences in the utility estimates. Additionally, questions could be asked at the beginning and near the end of the survey in order to find out whether this alternative approach may have influenced the results.
real life setting, in which the customer could explore beyond the stimulus website. This might have biased the results, since the respondents were unable to look up any additional information about any one of the attributes.
A fourth limitation of this research is that the incentive in return for the customers’ disclosure of personal information might have been vague, since nothing was communicated about the exact nature of the incentives. For instance, it was not communicated for what exact service or product the discount could be used or what the nature of the free newsletter would be. This may have affected the results since the participants did not know precisely what the exchange would be for the information requested. Further research would be crucial to find out whether the results of this investigation hold for incentives from which the exact value is communicated by the researcher.
5.4 Conclusion
preference for one of the attributes, if they are present on the webpage. Finally, customers who have a high score on privacy concerns are not as risk reluctant as one would expect.
7. Appendix
Appendix 1: Exemplary choice set Appendix 2: Factor Analyses Results KMO Barlett’s test ofSphericity Communalities above ,4 Number of factors explained Variance Privacy
concerns ,808 231,495 (p= 0,000) All 1 67,29%
Appendix 3: Utility Estimates with al the P-values
Model one Model two Model three Model four
Attributes Utility P-value Utility P-value Utility P-value Utility P-value
Retailer reputation No reputation -0,2691 0,000* -0,2653 0,000* -0,2686 0,000* -0,269 0,000* Bol.com 0,3561 0,3599 0,3616 0,3604 Baleno -0,0869 -0,0946 -0,093 -0,0913 Incentives Subscription to free newsletter -0,3414 0,000* -0,3546 0,000* -0,3574 0,000* -0,3575 0,000* A 10% discount 0,3383 0,3446 0,3444 0,3433 A one-month trial free membership 0,0031 0,01 0,013 0,0141 Information requested High level of info requested -0,6019 0,000* -0,6056 0,000* -0,6087 0,000* -0,6042 0,000* Medium level of info requested 0,0016 -0,0026 0,0015 -0,0055 Low level of info requested 0,6002 0,6082 0,6072 0,6097 Seals of approval Thuiswinkelwaarborg 0,2617 0,000* 0,2559 0,000* 0,258 0,000* 0,2697 0,000* Webwinkelkeur 0,0909 0,0956 0,092 0,2048 No seal of approval -0,3526 -0,3514 -0,35 -0,4745 Privacy statement Elaborate privacy statement 0,37 0,000* 0,3782 0,000* 0,3817 0,000* 0,3815 0,000* Unelaborate privacy statement 0,1436 0,1446 0,1467 0,1494 No privacy statement -0,5137 -0,5228 -0,5284 -0,531 No-choice option -0,0125 0,000* -0,018 0,000* -0,037 0,000* -0,0387 0,000* PC*No reputation -0,2021 0,21 -0,2103 0,25 -0,2038 0,2 PC*Bol.com 0,2796 0,08*** 0,2821 0,084*** 0,2775 0,073***
PC*Baleno -0,1892 N/A -0,1648 N/A -0,165 N/A
PC*Newsletter -0,1758 0,00047* -0,1785 0,00048* -0,1721 0,0003*
PC*Discount 0,2711 0,11 0,2667 0,09*** 0,2618 0,076***
PC*Trial membership -0,0952 N/A -0,0882 N/A -0,0756 N/A
PC*High level of
information requested -0,6289 0,67 -0,6333 0,66 -0,6293 0,65
PC*Medium level of
information requested -0,0478 0,34 -0,0381 0,41 -0,0441 0,42
PC*Low level of
information requested 1,2164 N/A 1,2786 N/A 1,2831 N/A
PC*Thuiswinkelwaarbo
rg 0,305 0,28 0,3142 0,22 0,328 0,21
PC*Webwinkelkeur 0,0528 0,37 0,0511 0,39 0,1773 0,57
PC*No seal -0,3577 N/A -0,3653 N/A -0,1929 N/A
PC*Elaborate privacy
statement 0,3675 0,82 0,3721 0,83 0,3692 0,79
PC*Unelaborate
privacy statement 0,1909 0,32 0,1946 0,31 0,195 0,33
