The privacy paradox: the effects of personalisation in different service-category types on consumers’ willingness of self-disclosure
Master thesis
Student: Choi-kuen (Connie) Tang UVA-ID: 11175753 Supervisor: Dr. H. Güngör Amsterdam, 22nd June 2018 Final version Executive Programme in Management Studies – Digital Business Track Institute: University of Amsterdam, Faculty of Economics and Business
Page 2 of 49 Statement of Originality This document is written by Choi-kuen (Connie) Tang who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Page 3 of 49
Abstract
In the age of digital disruption, personalisation has been recognized by practitioners as one of the most essential skills that firms should focus on. Personalisation can help firms attract consumers and help them bypass their competitors. More importantly, personalisation at an individual level enhances the stickiness of the entire customer journey, which increases firms’ engagement with their customers and lock customers into the loyalty loop. Conversely, highly personalised services may seem intrusive to consumers. Despite all these advantages, consumers are facing a trade-off between losses and gains when they are asked by firms to provide personal information. Although they are concerned about the safety of their data, consumers continue to provide personal information to firms. This inconsistency has been recognised as the personalisation privacy paradox (Awad & Krishnan, 2006). This thesis focuses on a further exploration of the personalisation privacy paradox and compares the differences in the willingness of information disclosure between two service types: hedonic versus utilitarian. This thesis also addresses the fact that information disclosure intention and information disclosure behaviour are two distinct constructs. This thesis focused on the information disclosure intention construct. In addition, by applying the privacy calculus theory (Culnan & Armstrong, 1999), perceived benefits and perceived risks were tested as mediators and privacy concern was tested as moderator between both experimental groups, perceived benefits and perceived risks. A vignette study conveyed that the willingness of self-disclosure was higher in the hedonic group than the utilitarian group. Furthermore, the relationship between the utilitarian group and the willingness of self-disclosure was mediated by the perceived risk, which showed that an increase in perceived risk led to a decrease in the willingness of self-disclosure. Although the expected mediation effect of perceived benefit in the hedonic group was not statistically significant, an increase in perceived benefits did lead to a higher willingness of self-disclosure. However, this study did not found any evidence for a moderating effect of privacy concern. This could possibly be explained by the bounded rationality theory. The privacy-related decision-making process may be situational, and consumers may make their privacy decisions based on heuristics instead of entirely rationally. Keywords: personalisation, privacy paradox, privacy concern, self-disclosure, online service, perceived benefits, perceived risksPage 4 of 49 Table of Contents 1. Introduction ... 6 2. Literature review ... 8 2.1 Willingness of self-disclosure ... 8 2.2 Personalisation ... 9 2.3 The value of online personalisation ... 10 2.4 Hedonic versus utilitarian service ... 11 2.5 Personalisation privacy paradox ... 13 2.6 Perceived benefits and perceived risks ... 14 2.7 Utility maximization theory ... 15 2.8 Privacy perceptions ... 16 2.8.1 Privacy concern ... 17 2.8.2 Influencers on privacy concern ... 17 3. Conceptual framework ... 18 4. Methodology ... 19 4.1 Research design ... 20 4.2 Online experiment ... 20 4.3 Participants ... 21 4.4 Measures ... 22 4.5 Control variables ... 23 5. Results ... 24 5.1. Data preparation ... 24 5.1.1 Manipulation check ... 25 5.1.2 Homogeneity of variances check ... 26 5.2. Data analysis ... 26 5.2.1 Testing hypothesis 1 ... 26 5.2.2 Testing hypotheses 2 and 4 ... 28 5.2.3 Testing hypotheses 3 and 5 ... 30 5.2.4 Testing hypotheses 6a and 6b ... 32 6. Discussion and conclusion ... 33
Page 5 of 49 6.1 Discussion ... 33 6.2 Managerial implications ... 35 6.3 Limitations and future research ... 36 6.4 Conclusion ... 37 7. Bibliography ... 40 8. Appendix ... 45 8.1 Webpage utilitarian ... 45 8.2 Webpage hedonic ... 46 8.3 Online experimental vignette study – survey ... 47
Page 6 of 49
1. Introduction
Total Dutch e-commerce spending reached a level of 600 million euros in the first six months of 2017 (Central Agency for Statistics, 2017). In the age of digital disruption, consumers and companies are continuously challenged by the rapidly changing environment, especially since data analytics have become more important and advanced. Consumers are empowered by innovation and digitalization, and firms are driven to expand their data collection and analytical capabilities. On one hand, data analytics enable firms to gain valuable insights about their existing and potential consumers. On the other hand, consumers expect firms to provide better and yet more customized services or products which should better fit their personal needs. Web personalisation has frequently been discussed in the top 20 marketing and information-system journals (Salonen & Karjaluoto, 2016). Additionally, the customer journey has received an increasing amount of attention within the digital marketing field (Kannan & Li, 2016; Rawson, Duncan & Jones, 2013). These marketing facts are forcing many firms to attract their potential customers by providing an engaging customer experience at the right place and at the right time. (Mckinsey, 2009). Simultaneously, market developments such as mobile commerce and high demand for personalisation have been highlighted as essential trends which are expected to drive changes in the consumer landscape toward 2030 (Mckinsey, 2015). Mobile marketing, location-based targeting, personalisation and recommendations at the individual level are market factors which are driving e-commerce firms to expand and optimize their data capture, storage and analysis capabilities (Lamberton, Stephen, 2015). From an innovation perspective, digital disruption has fundamentally changed the ways firms operate and shape their digital landscape (Weill & Woerner, 2015). Additionally, disruptive technologies enable firms to shift their services and products even more toward personalisation at the individual level. The pressure for e-commerce firms to better understand the consumers’ privacy intention and behaviour is continually growing (Mckinsey, 2015). Technological trends such as the internet of things (Perrig, Szewczyk, Wen, Culler & Tygar, 2001) and cloud computing (Subashini & Kavitha, 2011) are forcingPage 7 of 49 firms and consumers to consider their data privacy carefully when they expose themselves in the world of Big Data. From a regulation perspective, the General Data Protection Regulation (GDPR) announced changes to the existing data privacy law starting May 2018. The purpose of these changes is to give EU citizens more control over the processing and usage of their personal data. The GDPR applies to all EU or non-EU companies which collect or process personal data of any EU citizens and monitor their behaviour, regardless of whether these data-processing activities take place in the EU or not. According to the GDPR, firms should further strengthen their data-protection capabilities by assigning a data protection officer. When a data breach occurs, firms should be able to notify of the breach within 72 hours. Furthermore, gathering data from cookies, such as clickstreams and surfing histories, requires the consent of consumers upfront; consumer opt-in should be requested in an intelligible and easily accessible form. The GDPR gives consumers the right to be forgotten once the data are not relevant to the original purpose of processing (eugdpr.org, 2018). Although firms are aware of the urgency of complying the GDPR, many of them are still struggling to understand how they can interpret, measure and monitor compliance (Mckinsey, 2017). From the survey taken by Mckinsey in 2017, only one of the 19 participants in Europe believed his or her firm would fully comply by May 2018. Surprisingly, even though consumers are concerned about the safety of their personal data, they seem to behave inconsistently when they are requested to share their data (Norberg, Horne, Horne, 2007, Awad and Krishnan, 2006). As described in other research, consumers face a trade-off between the benefits and risks of giving their personal data to a firm. Several academic studies have referred to this trade-off as the personalisation privacy paradox (Awad & Krishnan, 2006, Dinev and Hart, 2006). Academic researchers have applied this phenomenon to explain why consumers are willing to share their personal information in different contexts, for instance, location-aware marketing and mobile commerce (Lee, Rha, 2016, Xu, Luo, Carrolla, Rosson, 2011). There is still a lack of research which applies this phenomenon in the context of online services (Lee, Rha, 2016). By taking a closer look at the personalisation privacy paradox, this study examined the relationship between
Page 8 of 49 personalisation in services (hedonic versus utilitarian) and the willingness of customers to share their personal information, leading to the following research question: RQ: How does personalisation in different service-category types (hedonic versus utilitarian) influence the level of consumers’ willingness to share personal information in the Netherlands?
2. Literature review
In an attempt to answer the research question, the reasons why people share personal data and when people do share personal information should be carefully investigated and explained. The first part of this literature review aims to define the factors that cause people concern about sharing their personal data. The second part proposes to establish the relationship between personalised service and the willingness of self-disclosure (WSD) by explaining why and when people tend to share personal data.2.1 Willingness of self-disclosure
In their book Self-disclosure, Privacy and the Internet, Joinson and Paine (2007) defined self-disclosure as “the telling of the previously unknown so that it becomes shared knowledge, the process of making oneself known to others.” This knowledge-sharing process happens between individuals, within groups or between an individual and an organization. More importantly, it has a variety of purposes depending on the context in which disclosure occurs. This thesis uses the term “personal information disclosure” to describe any form of information about the self that an individual provides. In the context of e-commerce, self-disclosure is not only an outcome of communication encounter, but also a product and process of dynamic interaction. Self- disclosure can build trust in romantic and friendship-based relationships. Disclosure can enhance the trust between group members (Joinson & Paine, 2007). Additionally, previous research has supported the existence of inconsistencyPage 9 of 49 between consumers’ intention to share personal information and their actual behaviour when they are asked to share personal information (Norberg, Horne, Horne, 2007). The two are recognized as the behavioural intention of self-disclosure and the actual behaviour of self-disclosure (Norberg, Horne, Horne, 2007). This paper focuses on the behavioural intention of self-disclosure. Personal information can be divided into market-level information and individual-specific information (Phelps, Nowak, Ferrell, 2000). Market-level information reflects generalized characteristics of a customer segment. In contrast, individual-specific information reveals identifiable data about an individual, for instance, age, interests, or shopping preferences. The level of consumer concern with individual-specific information is higher than with market-level information (Phelps, Nowak, Ferrell, 2000). Phelps, Nowak, and Ferrell (2000) found significant support for the theory that consumers are more willing to provide demographic and lifestyle information than financial, purchase-related and personal-identifier information.
2.2 Personalisation
In the context of e-business, personalisation refers to tailoring information and recommending products and services based on specific consumer characteristics (for instance, browsing or purchasing preferences) prior to the beginning of a customer’s search journey. It is an effective marketing strategy to target customers at the individual level, which provides various advantages to customers such as efficiency, convenience, individualisation, and hospitality (Chellappa, Sin, 2005). Personalisation is a critical value driver of a new form of e-commerce known as ubiquitous commerce. In the context of ubiquitous commerce, firms should aim to provide a higher degree of personalisation, which can provide additional benefits and values to customers with anything and to anyone, at any time and in any place (Sheng, Nah, Siau, 2008). In the rapidly changing digital environment, interaction with customers has become more personal. Personalisation has been shifting more toward an individual level in which contextual interaction plays an essential role in influencing consumers’ decision-makingPage 10 of 49 process. With the increasing popularity of mobile commerce, location and geography are important predictors of the consumer behaviour (Kannan and Li, 2016). Despite all these positive effects of personalisation, empirical evidence has shown that personalized ads may cause vulnerable feelings (Aquirre, Mahr, Grewal, de Ruyter and Wetzels, 2015) and appear intrusive to consumers (Doorn and Hoekstra, 2013). Consumers may perceive personalisation as a threat to their privacy (Aquirre, Mahr, Grewal, de Ruyter and Wetzels, 2015). Doorn and Hoekstra (2013) showed that a high degree of personalisation, such as adding identifying information, may increase the feeling of intrusiveness and, as a result, may decrease consumers’ buying intention. This negative effect of personalised advertising can be partly compensated by a good fulfilment of consumers’ current needs. Personalisation could be seen as a double-edged sword: Personalized advertising may lead to higher purchase intentions, but also to greater perceived intrusiveness, with simultaneously a decrease in purchase intentions (Doorn and Hoekstra, 2013).
2.3 The value of online personalisation
Consumers find different values in different types of personalisation; the value of personalisation positively influences the use of personalisation (Chellappa, Sin, 2005). In the context of retailing, consumers can directly benefit from personalised service by giving their consent to firms to collect their data. A study based on travel-industry data has indicated that travellers value the direct benefits, for instance, better price comparisons and personalized travel-package recommendations, which offset their privacy concerns (Lee, Cranage, 2011). Despite the presence of privacy concerns, consumers gain positive attitudes toward adoption of personalised services due to the perceived value of personalisation (Lee, Cranage, 2011, Sheng, Nah, Siau, 2008). The perceived usefulness of personalised services will offset privacy concerns once privacy assurance has been addressed by the firm (Lee and Cranage, 2011). Lee and Cranage’s results (2011) indicate that the assumption thatPage 11 of 49 personalised service triggers privacy concerns does not always hold. Consumers are willing to provide their personal data to gain a better quality of service. The value of online personalisation lies in the fit that a product or service provides to the customer and the convenience of having it delivered in a proactive way (Chellappa, Sin, 2005). Both perceived value and service quality are drivers of customer satisfaction (McDougall, Levesque, 2000). Online service quality could be measured by the following dimensions: (1) information availability and content, (2) usability, (3) privacy and security, (4) graphic style and (5) fulfilment (Parasuraman and Malhotra, 2002). Privacy perceptions influence customers’ overall perceptions of quality and value and their loyalty intentions within the online environment (Parasuraman, Zeithaml, Malhotra, 2005). Bauer, Falk and Hammerschmidt (2006) emphasized that the hedonics dimension (for instance, entertainment or fun) of e-services is critical to the value perceived by consumers using online services. Parasuraman, Zeithaml and Malhotra (2005) highlighted the difference in perceived value between consumers who are seeking hedonic benefits from a website and consumers who have a goal-directed shopping motivation as a topic for future research to examine. Hence, this research excludes service quality and focuses only on the value of personalisation. Customer characteristics such as consumer involvement, self-efficacy, consumer trust and technology readiness have a significant impact on the perceived value of adopting personalised service in the context of mobile marketing (Lee, Rha, 2016). Consumers who have high involvement and self-efficacy, who trust a service provider, and who have less technology insecurity perceived a higher value for adopting a personalised service. As well, Lee and Rha (2016) identified four consumer groups, each of which perceived a different level of benefits and risks in personalisation.
2.4 Hedonic versus utilitarian service
Once consumers begin the customer journey which has been shaped by online marketers, their consumption experience has two fundamental dimensional values, hedonic andPage 12 of 49 utilitarian, which cause two distinctive outcomes: experiential and instrumental (Babin, Darden, Griffin, 1994). Consumers are driven by different kinds of motivations before they search the internet, or in a local store, and purchase either online or offline. Consumers are driven by what they want or what they need. These two constructs belong to the terms hedonism and utilitarianism. Services and products are multidimensional (Voss, Spangenberg, Grohmann, 2003). As identified by Voss, Spangenberg, and Grohmann (2003), brand and product attitudes are complex and multidimensional. In a central route processing model, a two-dimensional view replaced the traditional, unidimensional view of consumer attitudes. The hedonic and utilitarian constructs are two distinctive dimensions to measure the underlying attitudes of the consumer. The hedonic dimension derives from sensations or experiences in using the product. In contrast, the utilitarian dimension derives from functions performed by the product. Both dimensions are valid and reliable to measure consumer attitudes toward brands and product categories. Measures on these dimensions may provide valuable inputs for understanding consumer behaviour (Voss, Spangenberg, Grohmann 2003). Additionally, the choice-making process of the consumer is driven by utilitarian and hedonic considerations (Dhar, Wertenbroch, 2000). Services and products may be categorized based on their relative hedonic or utilitarian nature. Hedonic services or goods provide more experiential consumption: fun, pleasure or excitement, which are what people want. Utilitarian services or goods are primarily instrumental and functional, which are what people need (Dhar, Wertenbroch, 2000). The product-trial research of Kemp (1999) defined the distinction as follows: Hedonic products are those consumed primarily for affective or sensory gratification purposes, and functional products deliver more cognitively oriented benefits. Their results showed that the effect of emotional responses and arousal on trial of hedonic products is much higher and stronger than utilitarian products. A further comparison of these two types of online services led to the following hypothesis:
Page 13 of 49 H1: Compared to personalisation in utilitarian services, the willingness of self-disclosure for personalisation in hedonic services is higher.
2.5 Personalisation privacy paradox
The majority of research within social networking and e-commerce fields has attempted to find the underlying explanation for why consumers keep providing personal information even after they have mentioned that they are concerned about their privacy. Previous researchers have termed this inconsistency between the expressed concern and the actual behaviour of consumers the privacy paradox: Consumers claim to be very concerned about their privacy, but at the same time they do very little to protect their personal data. To gain a better understanding of this phenomenon, researchers have applied multiple theories aiming to explain the privacy paradox. One stream of researchers has explained this paradoxical behaviour from a rational perspective by arguing that consumers weigh the cost-benefit ratio of online information disclosure in a conscious and rational way. They balance the perceived benefits and perceived risks (Awad, Krishnan, 2006, Dinev, Hart, 2006, Simon, 1955). Another stream has questioned the rationality of consumers. According to them, consumers are not able to estimate the benefits and costs due to their limited information. The decision-making process of individuals is partially rational when privacy is involved (Acquisti, 2004, Acquisti, Grossklags, 2005), yet consumers tend to underestimate the costs of disclosing information (Acquisti, Grossklags, 2005). Despite privacy concerns, consumers continue to provide their personal information to firm due to the existence of the privacy paradox (Norberg, Horne, Horne, 2007). One of the theories regarding the privacy paradox which has been used to explain this inconsistency between the intention and the behaviour of self-disclosure is known as the privacy calculus theory (Culnan and Armstrong, 1999). The privacy calculus theory describes consumers’ decision-making process as rational. Consumers make a valuation of gains versus losses. When consumers perceive more benefits than risks, they are willing to accept those risksPage 14 of 49 and provide personal information (Culnan and Armstrong, 1999, Dinev and Hart, 2006, Awad, Krishnan, 2006, Norberg, Horne, Horne, 2007, Xu, Luo, Carrolla, Rosson, 2011, Chellappa, Sin, 2005). However, it is assumed that there are other factors beyond perceived benefits and risks that influence the privacy paradox, for instance, privacy concerns and trust (Norberg, Horne, Horne, 2007). The privacy calculus is situation-specific. The decision depends on situational factors and situation-specific concerns. More importantly, the results of this valuation may change in different situations (Li, Sarathy and Xu, 2015).
2.6 Perceived benefits and perceived risks
In the context of location-based services, Xu, Teo, Tan and Agarwal (2009) showed that personalisation could be recognized as one of the essential dimensions of perceived benefits. Consumers are willing to provide their personal information in exchange for personalised services or information access. More importantly, the perceived benefits could be monetary or non-monetary benefits, for instance, convenience from online personalisation. Consumers tend to be willing to give up some degree of privacy for the potential benefits of personalised service (Xu, Teo, Tan, Agarwal, 2009). Additionally, Chellappa and Sin (2005) attempted to investigate the relationships between the likelihood of adopting personalised services, the value of personalisation, privacy concerns and trust-building factors. Within their research, the likelihood of adopting personalised services served as the dependent variable, and the value of personalisation, privacy concerns and trust-building factors were the predictors. Their findings showed that the value of personalised service was independent of privacy concerns while positively related to the likelihood of adopting personalised online service. Although the value of personalised service and privacy concerns were independent, Chellappa and Sin (2005) indicated that the value of personalisation to consumers was significantly more influential than consumers’ concern for privacy in predicting the likelihood of adopting personalised services.Page 15 of 49 In general, risks are uncertainties about potential negative outcomes and unwanted behaviour of the other party that would result in losses (Xu, Dinev, Smith and Hart, 2008). Academic papers have argued the relationship between the constructs of privacy concerns and perceived risk of privacy disclosure from two sets of beliefs: general beliefs and situation-specific beliefs. The research of Norberg, Horne, Horne (2007) took a closer look at the discrepancy between what people tend to say and what they tend to do regarding information sharing. Their research found that the level of an individual’s actual information disclosure significantly exceeded their level of intention to disclose information. Norberg, Horne, Horne (2007) challenged the idea that trust and privacy concern equally influence the privacy paradox. Instead, their research showed that trust does not influence the intention of self-disclosure and privacy concern does. Their work showed that perceived risk had a negative impact on the intention of self-disclosure, but less impact on the actual disclosure behaviour. As well, the difference in the level of information provided in all information categories (personally identifying, financial, preferences, demographics) may also influence actual self-disclosure behaviour (Norberg, Horne, Horne, 2007). Based on the findings regarding the perceived benefits and perceived risks, the following hypotheses were derived for this research: H2: Compared to personalisation in utilitarian services, personalisation in hedonic services will generate higher perceived benefits of information disclosure. H3: Compared to personalisation in hedonic services, personalisation in utilitarian services will generate higher perceived risks of information disclosure.
2.7 Utility maximization theory
From an economics perspective, when consumers make decisions about a certain level of risks, they face a trade-off between benefits and costs. During the decision-making process, consumers strive to maximize their benefits and reduce their costs, as stated by the utilityPage 16 of 49 maximization theory (Awad, Krishnan, 2006, Dinev, Hart, 2006, Simon, 1959). This theory assumes that consumers’ decision-making processes are rational and they aim to benefit maximally when they make choices. From this perspective, WSD is the outcome of the perceived value, which is
calculated as the perceived benefits of self-disclosure minus the perceived costs of self-disclosure. When the perceived benefits associated with disclosing personal information exceed the costs associated with disclosing personal information, consumers tend to be more willing to overcome their privacy concerns (Unni, Harmon, 2015). This research paper seeks to answer the research question by applying the utility maximization theory. The following hypotheses are derived based on this theory: H4: The level of perceived benefits of information disclosure is positively related to the willingness of self-disclosure. H5: The level of perceived risks of information disclosure is negatively related to the willingness of self-disclosure.
2.8 Privacy perceptions
Privacy perception varies among different cultures, and different cultures perceive different concerns due to their cultural values (Bellman, Johnson, Kobrin & Lohse, 2004). Bellman, Johnson, Kobrin and Lohse (2004) examined the differences in internet privacy concerns. One of their critical findings was that cultural values are associated with differences in privacy concerns. Cultural values are a set of beliefs that strongly influence the attitudes and behaviour of a population (Bellman, Johnson, Kobrin & Lohse, 2010). One of the well-known concepts developed by Hofstede (1984) has been used widely in academic and industry research. Hofstede divided cultural dimensions into power distance, individualism versus collectivism, masculinity versus femininity, long-term versus short-term orientation, indulgence versus restraint and uncertainty avoidance. Measured on the cultural dimensions of Hofstede, Dutch society scores very high in individualism, long-termPage 17 of 49 orientation and indulgence. Dutch individuals are expected to take care of themselves, orient pragmatically, and enjoy life, have fun and spend money on leisure time.
2.8.1 Privacy concern
The term “privacy” has been defined by several scholars based on different dimensions. Privacy could be defined as the right to be left alone (Dinev, Xu, Smith & Hart, 2013). In the context of consumer data, information privacy is the most frequently discussed term by scholars. Information privacy refers to the right to determine how information about oneself is communicated to others (Lowry, Cao & Everard, 2011). In the digital era, firms are finding ways to gather more customers' personal data and gain valuable insights regarding their customers (Awad and Krishnan, 2006). Consumers are concerned about the safety of their personal data once they have given consent to a firm to gather it. As a result, privacy concern influences the willingness of consumers to share personal information (Awad & Krishnan, 2006, Lee, Cranage,2011). Research within this area has shown that consumers are willing to pay a premium for websites that are privacy-protected (Tsai, Egelman, Cranor, Acquisti, 2011). As well, consumers who most value information-transparency features are the consumers who are least willing to be profiled online (Awad & Krishnan, 2006).2.8.2 Influencers on privacy concern
Based on the principles of the Federal Trade Commission, control over the collection of information, awareness of information collecting and usage of information beyond the original transaction strongly influence the quality of consumers’ experience and their privacy concerns (Sheehan, Hoy, 2000). Sheehan and Hoy (2000) highlighted two additional factors that contribute strongly to privacy concerns. These factors are the control that consumers could have by providing choice and the costs of the exchange that could occur by giving compensation.Page 18 of 49
Also, the study of Xu, Teo, Tan and Agarwal (2012) showed that perceived control has a mediating effect on the relationship between privacy concern and a privacy-assurance approach. Further, their study
found significant support for interaction effects of the three privacy-assurance approaches. Those approaches are individual self-protection, industry self-regulation and government legislation. The level of perceived consumer control is higher with individual self-protection. Consumers do not perceive relying on industry regulators as a robust approach (Xu, Teo, Tan, Agarwal, 2012). Privacy-assurance practices positively influence the willingness of consumers to share personal information by enhancing their trust and reducing their perception of risks (Lee, Cranage, 2011). Awad and Krishnan tested the effect of prior privacy-invasion experience on the level of willingness to share personal data. Their findings show that prior privacy invasion does lower the willingness share personal data in the context of personalised ads, but this does not hold in the context of personalised service because consumers may perceive different values (Awad, Krishnan, 2006). By taking a comprehensive view of the effects of privacy concern on the willingness of information disclosure, the following set of hypotheses were derived: H6a: Compared to personalisation in utilitarian service, personalisation in hedonic service will generate higher perceived benefits of information disclosure for people with low privacy concerns. H6b: Compared to personalisation in hedonic service, personalisation in utilitarian service will generate higher perceived risks of information disclosure for people with high privacy concerns.
3. Conceptual framework
This research was based on the expectation that personalisation in the hedonic service would generate higher willingness of self-disclosure compared to personalisation in utilitarian service. This main effect is mediated by the perceived benefits and risks. Privacy concern will moderate relationships between personalisation and perceived benefits inPage 19 of 49 hedonic service, and personalisation and perceived risks in utilitarian service. The conceptual model illustrated in Figure 1 serves as a summary of all hypotheses and their related effects as described above.
Figure 1. Conceptual Framework
4. Methodology
The first part of this chapter describes the research design, data collection and procedure which was used throughout this research. Subsequently, the measurements of the variables and control variables are discussed. The second part of the chapter describes further the sampling and data analysis techniques. Willingness of self-disclosure Perceived benefits Privacy paradox Perceived risks Hedonic service Value of Personalisation H2 Utilitarian service H3 H4 H5 H1 + + - + Privacy concern H6a A H6b APage 20 of 49
4.1 Research design
This research relies on an experimental vignette design, followed by an online questionnaire. By using short descriptions of a situation that are shown to respondents within the survey, a vignette study aims to elicit the judgement of the participants about these conditions (Atzmüller, Steiner, 2010). The main advantage that comes with an online experimental vignette study is rapid acquisition of respondents and respondents’ ability to fill in the survey in a natural environment, which can avoid laboratorial effects and researchers’ influence on the outcome. This experiment contains two conditions (personalised hedonic service, versus personalised utilitarian service), followed by an online questionnaire containing 29 questions. First, each participant was randomly assigned to one of the two conditions and was asked to respond to an upfront question about whether they would sign up for the website. The upfront question aimed to indicate the main effect of whether people were willing to provide personal information. Second, the indirect effects were tested based on two mediating variables, perceived benefits and perceived risks, by asking the participants six related questions. Third, another six questions regarding privacy concerns were presented to the participants. The last part of the survey contained six questions concerning the control variables and four asking about participants’ demographics. Participants could complete the questionnaire by entering the Qualtrics online survey environment.4.2 Online experiment
The online experiment relied on a between-subject design with two manipulations. Participants were randomly assigned to one of the two conditions: a web page with a hedonic service (Appendix 8.2) versus a web page with a utilitarian service (Appendix 8.1). Each participant was asked whether they would sign up for the web page by providing their name and email address. This experiment adapted the definition previously used by Drolet, Williams and Lau-Gesk (2007) to define each service category. The hedonic dimension reliesPage 21 of 49 on how the service or usage of the product makes someone feel, versus the utilitarian dimension, which relies on careful consideration of product or service attributes. In the work of Drolet, Williams and Lau-Gesk (2007), a pre-test was conducted among ten products and services. Their results showed that investment services and a pain reliever both belong to the utilitarian category, while greeting cards and cologne are relatively hedonic. Based on the suggestions of Drolet, Williams and Lau-Gesk (2007), two separate web pages were created: one with a hedonic service (a web page which provides service to customize your perfume, Appendix 8.2), and other with a utilitarian service (a web page which provides customized investments advice, Appendix 8.1). Aldas-Manzano, Ruiz-Mafe, Sanz-Blas and Lassala-Navarré (2011) showed that there is a positive interaction effect between perceived risks and trust. In a risky situation, for example, an online banking environment, the need for trust by consumers increases. Therefore, additional assurances of privacy and a Verisign security label were added to both web pages to control the possible effect of trust on the perceived risks in this online experiment. See Appendix 8.1 and Appendix 8.2.
4.3 Participants
This research applied a non-probability sampling technique (Dooley, 2001); methods such as self-selection, snowball and convenience sampling were used to collect the required data. The data for this research was gathered via the Qualtrics platform, Facebook, Twitter and LinkedIn. Two web pages with different service types were randomly assigned to participants via the survey link, which was available on these platforms. Furthermore, the QR code for entering the survey was printed and distributed at the University of Amsterdam and local supermarkets. As well, participants were asked to disseminate the survey link to their own social, work and education networks. According to Statistics Netherlands, in the year 2011, 69% of Dutch e-shoppers between the ages of 16 - 75 years bought products or services for personal use online. Therefore, the population of interest for this research was defined as people who are living in thePage 22 of 49 Netherlands. To ensure this, participants in the survey were asked to indicate whether they live in the Netherlands by answering an additional yes/no question.
4.4 Measures
An extensive overview of all items to measure the below constructs appears in Appendix 8.3. Willingness of self-disclosure (WSD) This study adapted the items previously used by Awad and Krishnan (2016) to measure WSD. The wording of the questions was modified to fit the context of this research. This construct is measured by a five-point Likert-scaled item ranging from “1 = definitely not” to “5 = definitely would”. Consumers’ value of personalised service (PER) Consumers’ likelihood of using a personalised service is strongly influenced by the value of personalised service (PER; Chellappa & Sin, 2005). Chellappa and Sin (2005) developed six items related to this construct that are tested and used throughout their research. Due to the similarity of the research, this study adapted three of the six items from Chellappa and Sin (2005) to measure consumers’ PER. Each of the items is measured using a seven-point Likert scale ranging from "1 = strongly disagree" to "7 = strongly agree". Perceived benefit (BEN) and Perceived risk (RIS) This research adapted one of the personalisation dimensions and the three items previously used by Xu, Teo, Bernard, Tan and Agarwal (2014) to measure the perceived benefits of information disclosure (BEN). In line with previous research of Xu, Teo, Bernard, Tan and Agarwal (2014), a seven-point Likert scale was used ("1 = Strongly Disagree" and "7 = Strongly Agree"). Regarding the perceived risks, prior research (Xu, Teo, Bernard, Tan and Agarwal, 2014; Unni, Harmon, 2015) identified the perceived risks of information disclosure (RIS) as a single-dimensional construct and defined this as the expectation of losses associated with the release of personal information. By adapting the three items of Xu, Teo, Bernard, Tan and Agarwal, 2014, and Unni, the perceived risks of information disclosurePage 23 of 49 (RIS) was measured by a 7-point Likert-type scale ranging from "1 = strongly disagree" to "7 = strongly agree". Privacy concern (CON) Concern for information privacy (CON) is a multidimensional construct (Awad, Krishnan, 2006). Malhotra, Kim and Agarwal (2004) tested this construct based on three dimensions: collection, control and awareness of privacy practices. Further, Malhotra, Kim and Agarwal (2004) referred to information privacy as "the claim of individuals, groups or institutions to determine for themselves when, how and to what extent information about them is communicated to others" (Westin 1967, p. 7). In the context of information privacy, concerns relate to an individual’s subjective view of fairness (Malhotra, Kim and Agarwal, 2004). This paper adapted the same constructs and six items used by Malhotra, Kim and Agarwal (2004) to measure this construct in the same context. Therefore, CON was measured by a seven-point Likert-scaled, ranging from "1 = strong disagree" to "7 = strongly agree”.
4.5 Control variables
This research adapted the same items and scales used by Xu, Teo, Bernard, Tan and Agarwal (2014) to define the following control variables: 1. Prior experience with online service (PES), as measured by a 5-point Likert-type scale (1 = never; 2 = one to three times; 3 = four to six times; 4 = seven to nine times; 5 = 10 times or more). 2. Previous privacy experience (PPE) as measured by a 7-point Likert-type scale (1 = none at all; 7 = very often). 3. Personal innovativeness (PIV) as measured by a 7-point Likert-type scale (1 = strongly disagree; 7 = strongly agree). By applying the control variables used by Awad and Krishnan (2006), the same consumer demographics, including age, gender and education background, were defined as additional control variables.Page 24 of 49
5. Results
This section presents the data that were collected and analysed. The first part contains the data cleaning procedure. The second part of this section indicates the analyses and the corresponding results that test the hypotheses and research question as stated above.5.1. Data preparation
The total number of participants was 206. Twenty-six of them did not fill in any question; therefore, those were removed from the dataset. Seven respondents partially filled in the survey questions. This research did not remove these missing values, but instead they were coded with "999" for further analysis to avoid reducing the statistical power of the dataset. The final dataset of this thesis research contained 180 participants. This sample size should be representative for conducting online research (Dooley, 2001). The age of the participants varied between 18 and 75 years old; 56% were male and 44% were female. Concerning educational level, 62% of the participants had achieved a university or higher vocational education level, 17% had achieved an intermediate vocational education level, and 21% were below the intermediate vocational education level. All 180 participants confirmed that they lived in the Netherlands. Subsequently, the data were checked for counter-indicative variables. In addition, dummy variables were created by coding each group as zero and one. Finally, a reliability analysis was performed for all constructs. The results of the reliability test are presented in Table 1, including a general overview of the constructs indicating the means, standard deviations and correlations.Page 25 of 49
5.1.1 Manipulation check
This research adapted the questions previously used by Drolet, Williams and Lau-Gesk (2007) to conduct manipulation checks. Participants within each condition were asked to respond to two statements: “The decision to buy a brand in this product category is mainly logical or objective” (Logicaltot), and “The decision to buy a brand in this product category is based a lot on feeling” (Emotiontot). Responses were recorded on a seven-point Likert-type scale anchored by "1 = strongly disagree" to "7 = strongly agree". By conducting a one-way ANOVA, participants who perceived the website as hedonic were expected to have a higher mean score for the statement regarding feeling than participants who perceived the website as utilitarian. The results graphed in Figure 2 are consistent with the research of Drolet, Williams and Lau-Gesk (2007). Investment service was perceived as a relatively more utilitarian category and perfume was perceived as a relatively more hedonic category.
Figure 2: Manipulations – means plot Tabel 1. Mean, Standard Deviations, Correlations
Variables Mean SD Age Education Gender Country WSDtot PERtot BENtot RIStot CONtot PEStot PPEtot PIVtot Utilitarian group Age 36.55 13.788 Education 4.56 1.424 -.170* Gender 1.45 0.498 -0.123 0.023 Country 1.01 0.076 -0.072 0.077 -0.068 WSDtot 2.7401 1.05001 -0.09 0.12 -0.07 .168* PERtot 3.5446 1.14564 0.052 -.154* 0.086 0.12 -.257** (0.611) BENtot 3.3873 1.33514 .206** -.228** -0.112 -0.022 -.479** .390** (0.856) RIStot 3.1773 1.31996 -0.021 -0.032 -0.009 0.106 .300** 0.107 -0.068 (0.832) CONtot 2.9977 1.18125 -0.072 -0.055 -0.087 0.143 .202** 0.065 0.069 .548** (0.853) PEStot 3.6705 1.32993 0.057 0.12 0.126 0.076 0.071 0.002 -0.053 0.092 0.094 PPEtot 3.1561 1.01668 -0.047 0.048 0.011 -0.049 -0.052 0.220** 0.007 -0.073 -.149* .193* PIVtot 3.7071 1.3431 0.069 -0.06 0.129 0.055 -.196** 0.263** .195* 0.097 0.12 0.127 0.002 (0.839) Utilitarian group 0.47 0.501 -0.071 0.148 0.009 -0.071 -.181* .154* 0.081 -.272**-0.136 -0.084 0.109 0.004
* Correlation is significant at the 0.05 level (2-tailed). ** Correlation is significant at the 0.01 level (2-tailed).
Page 26 of 49
5.1.2 Homogeneity of variances check
Before testing the hypotheses, it is essential to check the variances among both groups by performing a Levene test. The significance value of this test should be higher than 0.05. The results of the Levene’s test was not statistically significant (see Table 2), which indicated that equal variances among these two groups could be assumed. Further analysis could be performed.5.2. Data analysis
After conducting both checks and preparing the data, the hypotheses could be tested by analysing these data in SPSS. Throughout the entire data analysis stage, a significance level of .05 was used to test all hypotheses. Following a description of the data analysis, the research question is answered and a conclusion is drawn based on the findings of the analyses.5.2.1 Testing hypothesis 1
To test the H1 hypothesis, a one-way ANOVA was conducted in SPSS. The hedonic group was coded as the reference group and the utilitarian group served as the independent variable, with age, education, gender, PES, PPE and PIV as control variables. The results showed that there was a statistically significant effect of personalisation on WSD, F (1,175) = 5.90, p < .05. Also, WSD in the hedonic group was higher than WSD in the utilitarian group. See Figure 3.Tabel 2. Test of Homogeneity of Variances
Variable: WSD
Levene Statistic df1 df2 Sig.
Page 27 of 49 Figure 3 Willingness of self-disclosure- mean plot For a comparison of the difference between these two groups, an additional independent samples t-test was performed. The results of the test revealed that WSD in the hedonic group (hedonic = 0) was significantly higher compared to the utilitarian group (utilitarian = 1, p = .016). See Table 3. To further analyse the ability of personalisation to predict WSD in both groups, a hierarchical multiple regression analysis was conducted. In the first step, six predictors were entered: PIV, PPE, PES, age, education and gender. This model is statistically significant, F (6, 154) = 2.437, p < 0.05, and explained 8.7% of variance in WSD. After entering the utilitarian group at step 2, the total variance explained by the model as a whole was 13.5%, F (7, 153) = 3.399, p < 0.05. The introduction of personalisation explained an additional 4.7% of variance in WSD after controlling PIV, PPE, PES, age, education and gender (R2 change = .047, F (1, 153) = 8.368, p < 0.05. In the final model, only the variable PIV (β = -.227, p < 0.05) and the Tabel 3. ANOVA SS DF MS F Sig. WSD 6.329 1 6.329 5.9 0.016 Error 187.717 175 1.073 Total 194.045 176
Page 28 of 49 utilitarian group (β = -.224, p < 0.05) were statistically significant. In other words, if WSD in the hedonic group (hedonic = 0) increases by 1.0, WSD in the utilitarian group (utilitarian = 1) will decrease by 0.224 units. Based on these findings, H1 is accepted: Compared to personalisation in a utilitarian service, WSD for personalisation in a hedonic service is higher. Table 4 Regression Model for WSD
5.2.2 Testing hypotheses 2 and 4
A mediation analysis using Hayes’ PROCESS Model 4 was conducted to test the H2 and H4 hypotheses. In this case, the utilitarian group was coded as the reference group, and the hedonic group served as the independent variable, with age, education, gender, PES, PPERegression Model for willingness of self-disclosing
R R Square R Square Change B SE β t
Step 1 .294a 0.087* PPEtot -0.077 0.079 -0.076 -0.969 PIVtot -0.177 0.061 -0.226 -2.888 PEStot 0.064 0.063 0.083 1.023 Age -0.006 0.006 -0.086 -1.085 Education 0.043 0.056 0.06 0.763 Gender -0.223 0.164 -0.107 -1.363 Step 2 .366b 0.134* 0.047* PPEtot -0.048 0.078 -0.048 -0.615 PIVtot -0.179 0.06 -0.227* -2.975 PEStot 0.04 0.062 0.051 0.646 Age -0.007 0.006 -0.093 -1.199 Education 0.068 0.056 0.096 1.224 Gender -0.228 0.16 -0.11 -1.427 Utilitarian group -0.462 0.16 -0.224* -2.893
Page 29 of 49 and PIV as control variables. First, the results represented by a1 revealed that on average the hedonic group (hedonic = 1) does not show a statistically significant difference in perceived benefits compared to the utilitarian group (utilitarian = 0). The effect of the hedonic group on perceived benefits, a1 = - 0.3740, see table 5, does not differ significantly from zero, t = -1.8475, p = 0.066, with a confidence interval from -.7739 to 0.0259. Therefore, H2 is rejected. Subsequently, the direct effect is c1 ʹ = 0.3382, see table 5, meaning that participants in the hedonic group who perceive the same level of benefits compared to the utilitarian group are estimated to be 0.3382 units higher in their WSD intention. This direct effect is statistically different from zero, t = 2.2992, p = 0.0229, with a 95% confidence interval from 0.0476 to 0.6288. The total effect of the hedonic group on WSD is c1 = 0.4622, see table 6, meaning that a participant in the hedonic group who perceives the same level of benefits as a participant in the utilitarian group is estimated to be 0.4622 units higher in their WSD intention. This total effect is statistically different from zero, t = 2.8927, p = 0.0044, or between 0.1465 and 0.7779 with a 95% confidence interval. Based on these outcomes, H4 is accepted. Table 5
Antecedent Coeff. SE p Coeff. SE p
Hedonic Group (X) a1 -0.374 0.2024 0.0666 c1’ 0.3382 0.1471 <.05 BENtot (M) --- --- --- b1 -0.3316 0.0581 <.05 constant i1 3.3771 0.7024 <.05 i2 4.5201 0.5416 <.05 Consequent BENtot (M) WSD (Y) R2 = .1459 R2 = .2869 F (7,153) = 3.7341, p <.05 F (8,152) = 7.6437, p <.05
Page 30 of 49 Table 6
5.2.3 Testing hypotheses 3 and 5
By examining the same mediation analysis as described in the previous section, this section describes the test of the H3 and H5 hypotheses. In this case, the hedonic group was coded as the reference group and the utilitarian group served as the independent variable, with age, education, gender, PES, PPE and PIV as control variables. The effect of the utilitarian group (utilitarian = 1) on perceived risks, a1 = -.6921, see table 7, indicated that a participant in the utilitarian group who differs by 1.0 from a participant in the hedonic group (hedonic = 0), is estimated to differ by a1 = -.6921 units on perceived risks. The sign of a1 is negative,meaning that on average, participants in the utilitarian group are estimated to be 0.6921 units lower in their perceived risks than participants in the hedonic group. This effect is statistically different from zero, t = -3.2590, p = 00014, with a 95% confident interval from -.1116 to -.2725. The effect of b1 = 0.2085, see table 7, indicates that participants in the utilitarian group and the hedonic group who differ by one unit of perceived risk are estimated to differ by b1 = 0.2085 units in WSD. The sign of b1 is positive, meaning that those in the utilitarian group are estimated to be 0.2085 units higher in their WSD. The effect is statistically different from zero, t = 3.5564, p = .0005, with a 95% confidence interval from .0927 to .3244. The direct effect of the utilitarian group is c1 ʹ = -0.3179, see table 7, meaning that participants in the utilitarian group who perceive the same level of risk as those in the hedonic group are
Effect SE p LLCI ULCI
Direct
effect c1’ 0.3382 0.1471 <.05 0.0476 0.6288
Total effect c1 0.4622 0.1598 <.05 0.1465 0.7779
Boot SE Boot LLCI Boot ULCI
Indirect
Page 31 of 49 estimated to be .3179 units lower in their WSD. This direct effect is statistically different from zero, t = -1.9957, p = 0.0478, with a 95% confidence interval from -0.6326 to -0.0032. The indirect effect of the utilitarian group is -0.1443, see table 8, meaning that on average participants in the utilitarian group are estimated to differ by -0.1443 units in their WSD compared to the participants in the hedonic group. In other words, a participant in the utilitarian group who perceived more risks compared to a participant in the hedonic group had a lower WSD. This indirect effect is statistically different from zero, as revealed by a 95% BC bootstrap interval entirely below zero (-0.3038 to -0.0466). The total effect of the utilitarian group on WSD is c1 = -0.4622, see table 8. Participants in the utilitarian group who perceived the same level of risk are estimated to differ by -0.4622 units in their WSD. The negative value means that participants in the utilitarian group who perceived a greater level of risks reported lower WSD. This effect is statistically different from zero, t = -2.8927, p = 0.0044, with a 95% confidence interval from -0.7779 to -0.1465. Based on these results, H3 and H5 are both accepted. Table 7
Antecedent Coeff. SE p Coeff. SE p
Utilitarian Group (X) a1 -0.6921 0.2124 <.05 c1’ -0.3179 0.1593 <.05 RIStot (M) --- --- --- b1 0.2085 0.0586 <.05 constant i1 3.6742 0.7234 <.05 i2 3.8625 0.5443 <.05 Consequent RIStot (M) WSD (Y) R2 = .0847 R2 = .2006 F (7,153) = 2.0236, p <.05 F (8,152) = 4.7672, p <.05
Page 32 of 49 Table 8
5.2.4 Testing hypotheses 6a and 6b
A moderated mediation analysis using Hayes’ PROCESS Model 8 was conducted. In this case, the hedonic group served as the independent variable, with the utilitarian group as the reference group, perceived benefits as mediating variable, WSD as the dependent variable and privacy concerns as moderating variable to test H6a. Age, education, gender, PES, PPE and PIV were set as the control variables. The results revealed that privacy concern does not have a statistically significant effect on the relationship between the hedonic group and the perceived benefits. Hence, H6a is rejected. Another moderated mediation analysis was conducted to test H6b. In this case, the utilitarian group was coded as the independent variable, the hedonic group served as the reference group, perceived risks as mediating variable, WSD as the dependent variable and privacy concern as moderating variable. Age, education, gender, PES, PPE and PIV were the control variables. The results showed that privacy concern does not have a statistically significant effect on the relationship between the utilitarian group and the perceived risks. H6b is rejected. Privacy concern does not have a moderating mediation effect on the relationships between personalisation service types, perceived benefits and perceived risks.Effect SE p LLCI ULCI
Direct
effect c1’ -0.3179 0.1593 <.05 -0.6326 -0.0032
Total effect c1 -0.4622 0.1598 <.05 -0.7779 -0.1465
Boot SE Boot LLCI Boot ULCI
Indirect
Page 33 of 49
6. Discussion and conclusion
The first part of this section discusses the findings and compares them with previous research. The second part presents the managerial implications of this research, followed by its limitations and recommendations for future research. The last part of this section contains the conclusion of this paper.6.1 Discussion
The goal of this study was to test what effect different service types have on WSD by applying the personalisation privacy paradox theory and the utility maximization theory. From an academic perspective, this study contributes to and extends other research that investigates the inconsistent behaviour of consumers, known as the personalisation privacy paradox. In addition, this study focuses mainly on the context of services due to a lack of existing academic research in this field. The effect of personalisation in two different service types on WSD was investigated (H1). Subsequently, the indirect effects through perceived benefits and perceived risks were tested (H2, H3, H4, H5), followed by tests of the potentially moderating effect of privacy concerns (H6a and H6b). First, the findings represented by H1 indicate that personalisation in hedonic services generates higher WSD than personalisation in utilitarian services. Second, this study did not find any statistical evidence to support H2, which means that the hedonic service did not generate more WSD through the perceived benefits. As supported by the statistical evidence for H4, perceived benefits increased WSD for the hedonic group compared to the utilitarian group. On the other hand, this study found that WSD for the utilitarian group was lower compared to the hedonic group through the perceived risks. The findings support H3 and H5, in line with the assumptions of the utility maximization theory explained by Awad, Krishnan (2006), Dinev, Hart (2006) and Simon (1959). As mentioned above, this theory assumes that consumers’ decision-making process is rational and relies on the cost-benefit trade-off (Awad, Krishnan, 2006, Dinev, Hart, 2006, Simon, 1959). In the hedonic group, the indirect effect was rejected. From the products and services perspective, this could be explained by framing effects. One of the most cited papers inPage 34 of 49 psychology describes how the framing of decision-making problems through hedonic or utilitarian dimensions causes significant changes in humans’ preferences when they make choices (Tversky, Kahneman, 1981). Some products or services could be framed as hedonic or utilitarian, which would influence consumers’ decision-making process. This framing effect with different product types (hedonic versus utilitarian) was used by Baker, Fang, and Luo (2014) to investigate the effectiveness of mobile ads. Their experiment showed evidence that the framing of products could be either hedonic or utilitarian. Also, the framing of product types had different impacts on the effectiveness of the mobile ads throughout the day (Baker, Fang, Luo, 2014). As well, preferences between hedonic and utilitarian services or products depend how these two consumables are presented. Consumers tend to rate hedonic services or products higher than comparable utilitarian services or products when the hedonic item is presented singly; but the utilitarian service or product tends to be chosen over the hedonic service or product when these two are presented together. Additionally, consumers tend to be willing to pay more in time for a hedonic service and more in money for a utilitarian service (Okada, 2005). Lastly, this study expected that privacy concern would moderate the relationship between service types and perceived benefits and risks. Although previous studies found evidence that privacy concerns influence the willingness of people to share their data, these findings contrasted with that expectation. From the behavioural economics perspective, the bounded rationality theory could be an explanation which is based on the opposite view of the utility maximization theory. As argued by other academic researchers, the bounded rationality theory assumes that the choice-making process is bounded, which is caused by the limits of rationality. Rationality could be bounded due to incomplete information about alternatives, complexity of the costs or other environmental constraints. These limits prevent actors from making a rational choice, and so their choice is partly based on intuition or heuristics (Simon, 1972, Kahneman, 2003). Acquisti and Grossklags (2003) showed that consumers’ decision-making process could be both rational and irrational due to the existence of present-bias. Privacy decision-making is only partially rational, because this is affected by misperceptions about benefits and costs (Acquisti & Grossklags, 2005).
Page 35 of 49 To summarize, this study showed that WSD in a personalised hedonic service was higher compared to a personalised utilitarian service. Personalisation in the utilitarian service was negatively correlated to WSD through perceived risks. In the hedonic group, the perceived benefit was positively related to WSD and the perceived benefit did not mediate the relationship between the personalised hedonic service and WSD. More importantly, consumers’ privacy decision-making process may not be entirely rational, which means that they may make their privacy decisions based on heuristics instead of rationality. In addition, the framing effect may influence consumers' decision-making process among these two types of consumables.
6.2 Managerial implications
From a managerial perspective, the results of this study provide insight for firms that have chosen to focus on personalised services. On one hand, those firms aim to collect more personal data about their existing or potential customers. On the other hand, they are challenged by customers' increasing awareness of data privacy. Nowadays, as technology enables firms to track the customer’s entire journey at an individual level, WSD from the customer side enables those firms to perform in-depth analyses of the individual’s purchasing-decision process (Anderl, Schumann, Kunz, 2016). Anderl, Schumann, Kunz (2016) confirmed the interaction effects among firm-initiated channels (display, retargeting, affiliate, email) and customer-initiated channels (direct type-in, branded search, generic search, price comparison) from a taxonomic rather than a last-click view of the entire customer journey. Consumers who click on firm-initiated channels, such as display or retargeting, tend to have a higher propensity to purchase. Conversely, switching from a branded to a generic consumer-initiated channel showed a decrease in purchase probability (Anderl, Schumann, Kunz, 2016). The findings of this study could provide additional insights to marketing managers on ways to increase the attractiveness of each touchpoint, especially in firm-initiated channels. Also, this study supports the possible influence of framing effects (Baker, Fang, Luo, 2014). These effects may be beneficial for firms during their marketing-communication activities with their existing or potential customers. Products or services could be framed as eitherPage 36 of 49 utilitarian or hedonic. Each framing could have differing effects on the click-through rates of advertisements (Baker, Fang, Luo, 2014). Marketers could influence consumers’ intentions by framing the products or services correctly. In a service-oriented industry, firms could apply this framing effect to communicate with their customers, which may influence their willingness of sharing personal data with firms. Finally, the results of this study provide practitioners with valuable insights about consumers’ privacy decision-making processes. Consumers’ purchasing behaviour has been shifting more toward mobile platforms, and this trend should not be overlooked (Lee, Rha, 2016, Xu, Luo, Carrolla, Rosson, 2011). The increasing need to optimize mobile advertising strategies at a contextual level is forcing firms to increase their data-analytical capabilities (Grewal, Bart, Spann, Zubcsek, 2016). Practitioners could optimize their mobile advertising strategies to engage each customer, throughout the customer journey, where this customer is located. Given the increasing popularity of gamification, organizations could apply game-like elements to their existing strategies, or service and system usage, to focus on engagement and enjoyment elements of the consumer experience (Hamari, Koivisto, Sarsa, 2014). Hamari, Koivisto, Sarsa (2014) suggested that practitioners adopt the significant positive effects of gamification on user motivation and engagement in the marketing context. Gamification could be considered as a potential marketing strategy (Hamari, Koivisto, Sarsa, 2014). Previous research showed that consumers may make their privacy decisions based on heuristics due to their bounded rationality. Consumers tend to use heuristics to guide their behaviour; environmental cues such as personal characteristics and design elements of the website generate additional feelings of trust. Using trust as a heuristic element may significantly shorten the consumers’ decision-making process on disclosure (Norberg, Horne, Horne, 2007).
6.3 Limitations and future research
Despite the relevance of these findings, there are several limitations which should be addressed. First, due to the cross-sectional roots of this research, the results only reflect aPage 37 of 49 snapshot of consumers’ willingness and perception toward privacy. Future research at a different time might produce different results. Second, based on the vignette design of this randomized experiment, in which participants were randomly assigned across two conditions of service category type, the results obtained in this experiment are under somewhat artificial circumstances and are not entirely generalizable to real-life settings of service use. On the other hand, the experimental design enabled greater control over other factors that may affect WSD in a real-life setting and allowed a focus solely on the effect that a hedonic versus utilitarian service type may have. Third, in case of privacy and sharing personal information, people’s actual behaviour may differ from their intentions (Norberg, Horne, and Horne, 2007), which means the findings in this study may vary from participants’ actual behaviour of self-disclosure. Research about actual behaviour compared to intention for the future is expected to provide complementary insights. Further, a representative sample provides the ability to generalize a conclusion over the population (Dooley, 2001). Future research could increase the sample size to increase the generalizability. Fourth, as the phenomenon of information privacy may be culturally dependent, this research framework can be expanded across different cultures. By looking at the age-related attitude differences toward hedonic versus utilitarian products, previous research showed that young-adult consumers (younger than age 25) had more positive attitudes toward hedonic products than toward utilitarian. In contrast, older consumers do not perceive differences among these two product categories (Drolet, Williams, Lau-Gesk, 2007). Therefore, future research exploring the possible effect of age on WSD may provide new insights.
6.4 Conclusion
Firms request their customers to provide personal information, with personalised service as a possible outcome of this commercial exchange. Consumers do not behave consistently when they are asked to disclose information. They tend to provide substantially more information for commercial exchanges than they said they would when asked prior to the commercial exchange (Norberg, Horne, Horne, 2007). Researchers have confirmed thisPage 38 of 49 inconsistency and referred to this phenomenon in several academic papers as the personalisation privacy paradox (Awad & Krishnan, 2006, Dinev and Hart, 2006). Consistent with previous research, the results of this study confirmed the existence of the personalisation privacy paradox (Norberg, Horne, Horne, 2007, Li, Sarathy & Xu, 2015). The valuation between perceived risks and perceived benefits influences consumers’ intention to disclose information. Despite the privacy concerns that consumers may have, the perceived benefit outweighs the perceived risk, which may cause individuals to share personal information. Additionally, as shown by the results of this study, personalisation in the hedonic services may generate higher WSD than personalisation in the utilitarian services. In this study, perceived risk did negatively influence WSD, and conversely, perceived benefit had a positive influence on WSD. More importantly, as shown by these results, the privacy calculus may be situation-specific. The decision depends on situational factors and situation-specific concerns (Li, Sarathy & Xu, 2015). In addition, individuals may make decisions regarding information disclosure based on heuristics (Norberg, Horne, Horne, 2007) rather than the purely rational perspective (Culnan and Armstrong, 1999). In general, consumers’ WSD contains two dimensions: behavioural intentions and actual behaviour. Consumers’ self-disclosure intention may differ from their actual information-disclosing behaviour (Norberg, Horne, Horne, 2007). Trust positively influences the actual behaviour of information disclosure, and perceived risk decreases the intention of information disclosure (Norberg, Horne, Horne, 2007). The valuation between perceived benefit and perceived risk may vary in different situations (Li, Sarathy & Xu, 2015). In the context of online services, the framing effect may help practitioners to influence consumer choice. Some services or products could be framed either as hedonic or utilitarian. Different framing of products or services (hedonic versus utilitarian) could have different effects on consumers’ decision-making process (Baker, Fang, Luo, 2014). As confirmed by other research, privacy concern influences WSD (Awad & Krishnan, 2006, Lee, Cranage,2011). Although this research was not able to confirm the moderating effect of privacy concern between service types, perceived risk and perceived
