Consumers are swamped with new products and services on daily basis due to the modern, multifaceted, omni-channel environment that we live in today. The retail industry changes very quickly and is considered to be extremely competitive. Therefore, it is crucial to create value for customers and aim for customer retention (Grewal, Roggeveen, and Nordfalt 2017; Pan and Zinkhan 2006). Technology enables retailers to enhance utilitarian and hedonic shopping values, which often result in an increase of retail patronage intentions (Pan and Zinkhan 2006). Retail patronage intentions are defined in this study as the subjective probability that an individual will purchase products from the store in the future again (Chiu, Lin, Sun, and Hsu 2009). Understanding patronage behaviour is of vital importance, since it enables retail managers to identify and target consumers that are most likely to purchase (Pan and Zinkhan 2006). Although the definition seems straightforward and the construct is perceived to be of great value to the business and academic world, most studies have ambiguously used or even misused the construct (e.g. Ganesh, Reynolds, Luckett, and Pomirleanu 2010; Ou, Abratt, and Dion 2006; Pan and Zinkhan 2006; Baker, Parasuraman, Grewal, and Voss 2002). Moreover, over the last decade retail patronage studies have predominantly focused on e-commerce environments because this phenomenon has had a major impact on the retail industry (e.g. Davari, Iyer, and Rokonuzzaman 2016; Ganesh, Reynolds, Luckett, and Pomirleanu, 2010; Kim, Fiore, and Lee 2007). However, the Internet of Things (IoT) promises to be even more disruptive for the industry (McKinsey 2015). As e-commerce has been devastating for many physical retailers, IoT has the potential to connect and compete with the online world (McKinsey 2015).

The IoT has become one of the most important topics of interest within the field of information technology (Balaji and Roy 2017). Due to its recent and still continuous establishment there is no clear consensus about a general definition. Nevertheless, there is a broad assent that the foundation of the IoT paradigm involves everyday objects that possess a technological component with sensing, identifying, networking and processing capabilities, which allow them to communicate with other services or objects (Whitmore, Agarwal, and Xu 2015). The IoT is identified as one of the top strategic technology trends that are expected to shape business opportunities and could potentially grow into a market worth $7.1 trillion by 2020 (Garner 2015; Wortmann and Flüchter 2015). The retail industry is one of the industries that could potentially benefit the most from the IoT according to McKinsey (2015), Accenture (2015) and Balaji and Roy (2017).

IoT applications contribute to retail environments in various ways. For instance, the German grocery retailer Dohle has introduced smart shopping carts, which display real-time product information and answer ad hoc queries. Stores of the American jewellery retailer BaubleBar have used interactive and sensor enabled displays, which offer tailor-made product information to customers (Balaji and Roy 2017). Concerning the global retail industry, it is estimated that the IoT could have an economic impact of $410 billion to $1.2 trillion per year in 2025. It is expected that no-checkout technology, also referred

(4)

4 to as automated checkout systems (not to be confused with self-checkout systems), will have the largest impact on the retail industry (McKinsey 2015). The technology enables customers to walk into a store grab the products they want and just leave the store. No checkout procedure is required and the bill is automatically paid via a personal account. The technology is able to register all movements of customers and products due to cameras, deep learning, algorithms and sensors. The checkout process could potentially be reduced in time between 40 and 88 percent and as a result, cashier costs could be cut up to 75 percent in 2025. This could all lead to global savings of $150 billion to $380 billion (McKinsey 2015). Amazon, Alibaba and Lawson have already opened pilot stores that have been equipped with no- checkout technology. From a consumer perspective, the technology is designed to increase convenience as most important utilitarian shopping value and enjoyment as most important hedonic shopping value.

In case of implementation, it could therefore positively affect retail patronage intentions. However, the development of existing and innovative (IoT) retail technologies is taking place very rapidly. As a result, retailers indicate having difficulties determining how these rather complex technologies fit within their strategy and even more importantly, are perceived by consumers (Inman and Nikolova 2017). This research investigated how one of the most promising IoT technologies for retailers, no-checkout technology, affects consumers’ retail patronage behaviour intentions.

The biggest threat to the success of IoT technologies adoption are privacy risks and concerns (Weber 2015; Atzori, Iera and Morabito 2010). These concerns mainly relate to data collection, user’s control over data and awareness of the use of data (Caron, Bosua, Maynard, and Ahmad 2016). In the context of no-checkout technology, retailers are for example able to detect that a particular customer considers buying a specific product, but eventually did not. Without being aware of it, retailers are subsequently able to retarget the customer with advertisements of the product once he or she left the shop. Likewise, retailers are able to track whereabouts, spending habits and consuming habits (Weber 2015).

Despite the rapid advance of the technology and the major privacy concerns, Nguyen and Simkin (2017) claimed it is remarkable that only a few marketing studies have been conducted in the area of the IoT. The authors indicated that various areas and challenges need be researched in order to determine the impact for the stakeholders. Gao and Bai (2014) also stressed the dearth of research in regard to understanding consumers perceptions of IoT technology and Evanschitzky, Iyer, Pillai, Kenning, and Schute (2015) called for the need for further research concerning understanding antecedents of customer acceptance of IoT technologies. More specific for the retail industry, Balaji and Roy (2017) and Madhani (2015) emphasised the lack of empirical research regarding customers’

evaluations and perceptions of IoT technology in the retail sector. Furthermore, Inman and Nikolova stressed that because no-checkout technology probably has such a tremendous impact on the future of the retail environment, it has led to a broad open research area. The authors emphasised the importance of future research into vital facets of marketing, such as consumer behaviour.

(5)

5 Following the gaps identified in the sections above, the following research question has been proposed:

How do an increase of convenience and enjoyment perceptions, with regard to the implementation of no-checkout technology, affect retail patronage intentions? And how are retail patronage intentions

and the relationships mentioned above, affected by privacy concerns of this type of technology?

The main objective of this study is to provide evidence for the relationship between important antecedents and retail patronage intention, which have not been researched before or have not been researched in the IoT environment. Therefore, this study responds to the call to perform empirical research regarding customers’ evaluations and perceptions of IoT technology in a retail context (e.g.

Balaji and Roy 2017; Inman and Nikolova 2017; Madhani 2015). In addition, this research aims to offer clarity to the use of the retail patronage construct, since it has been ambiguously used in previous studies (e.g. Baker et al. 2002; Jones and Reynolds 2006). The research also offers managerial relevance because it provides retailers with insights on how consumers perceive and evaluate no-checkout technology. Moreover, this study provides evidence whether the implementation of no-checkout technology could increase retail patronage intentions and what particular factors cause this effect.

2. Literature Review

2.1 Retail Patronage Intentions

Patronage behaviour has been part of retailing since the 1920s. However, Sheth (1981) was the first who introduced an economic theory devoted to explaining this type of behaviour. The patronage behaviour theory posited that patronage behaviour is a function of preference-behaviour discrepancy, triggered by four types of events: 1) socioeconomic setting (e.g. inflation); 2) personal setting (e.g. available time);

3) product setting (e.g. brand availability); and 4) in-store marketing (e.g. promotions). Nonetheless, the theory is not broadly supported by the literature stream and is not perceived as a solid foundation for explaining retail patronage behaviour. More remarkably, no commonly accepted theory that explains the core of patronage behaviour exists. Arnold, Handelman, and Tigert (1996) stated that consumer store choice models provide a basis for explaining retail patronage processes within the retail environment (e.g. Burke, Harlam, Kahn, and Lodish 1992; Keng and Ehrenberg 1984; Louviere and Gaeth 1987).

Due to the fact that these models do not commit to a solid and consensual theoretical background, they are wide ranging and even contradictory in some cases.

The lack of a solid theoretical base and unanimity becomes clearly visible when reviewing patronage studies from decades ago until present. For example, numerous studies lack a clear definition of the construct and in most cases, leave it up to the reader to interpret the exact meaning (e.g. Baker et al. 2002; Ou, Abratt, and Dion 2006; Pan and Zinkhan 2006; Ganesh, Reynolds, Luckett, and

(6)

6 Pomirleanu 2010). This is remarkable, as for the reason that authors used and perceived the construct very differently across studies (e.g. Baker et al. 2002; Jones and Reynolds 2006). For example, Osman (1993) had defined five dimensions, based on earlier studies, that measure loyalty patronage: 1) purchase percentage of a product category at the specific store; 2) frequency of visiting the store; 3) ratio of store ranking; 4) the tendency to shop at the store in the future; and 5) the willingness to recommend the store to their friends. The author defined loyalty behaviour as “the repeat purchase behaviour at a particular store for either the same products or any other products’’ (p. 135). Shim and Kotsiopulos (1992) defined patronage behaviour as store choice behaviour that represents a consumer’s preference for a specific store for purchasing goods and/or services. Pan and Zinkhan (2006) have conducted a meta-analysis on retail patronage behaviour. The authors claimed patronage behaviour consists of two dimensions; 1) store choice and 2) frequency of visit. Although, the meta-analysis of Pan and Zinkhan (2006) is widely accepted and often cited, it does not provide a clear definition of the construct, nor does it tap into how retail patronage behaviour relates to closely related constructs (e.g. loyalty and purchase intentions).

Patronage and loyalty behaviour are most frequently used as distinctive variables, but often measured with exactly the same measurement instruments. For example, Grewal, Baker, Levy, and Voss (2003) conducted research on store patronage intentions with an identical metric Sirohi, McLaughlin, and Wittink (1998) used for capturing store loyalty intentions. This is remarkable since various researchers perceive these constructs as stand-alone and isolated constructs. For example Jones and Reynolds (2006), perceive repatronage intentions, loyalty and repatronage anticipation as different constructs within their study. The authors defined repatronage intentions as “the likelihood that a customer will shop at a retail store again’ and loyalty as a ‘deeply held commitment to a particular retailer” (p.976).

The discrepancies above, clearly show how authors have used different approaches for defining patronage behaviour (e.g. by different key elements; ‘store choice’, ‘frequency’, ‘preference’, and

‘willingness to recommend’) and measuring the construct. The following citation of Seock (2009) describes the underlying issue of patronage behaviour studies in a flawless, but unintended way:

…the degree of store loyalty or patronage behaviour can be measured by using several variables, and past studies on store loyalty or patronage behaviour have used either one or a combination of variables for this measurement. (Seock 2009, p. 331)

As many other studies, the author clearly acknowledged the existence of two different constructs and multiple ways for measuring the constructs, but subsequently perceived the construct to be identical and has not provided any argumentation on the substantial differences between the constructs. Consequently, results and measurement instruments of loyalty oriented studies have been used to support and explain patronage behaviour and vice versa. It can be concluded that previous studies have used the seemingly closely related constructs loyalty and patronage behaviour on one hand as separate constructs, claiming to measure different types of behaviour, and on the other hand interchangeably, measuring the same

(7)

7 type of behaviour. Moreover, it can be stated that researchers have lost track of the proper meaning of each construct and the fundamental differences. Studies on loyalty and patronage behaviour have often used various dimensions for explaining and measuring the different constructs, while often neglecting key elements that are stressed to be of great importance by others (Pan and Zinkhan 2006; Seock 2009).

Therefore, this study first explains the core elements of loyalty and patronage behaviour in order to define the critical differences and determine the proper use of retail patronage intentions in the remainder of this study.

As described, various authors argued that loyalty represents retail patronage behaviour and vice versa (e.g. Cole and Clow 2011; Hozier and Robles 1985). Critics, however, posited that loyalty behaviour is a rather complex construct and for example emphasise that patronage behaviour primarily captures consumers’ repurchase intentions (Davari, Iyer, and Rokonuzzaman 2016). Davari, Iyer, and Rokonuzzaman (2016) stated that patronage intentions cover the willingness to purchase at the store but do not touch upon the more complex psychological elements that comprise loyalty. The authors claimed, loyalty is a multifaceted phenomenon that needs time to develop since it indicates the resilience of a relationship between consumer and retailer. However, Mägi (2003) stated that the degree of store loyalty is influenced by the overall patronage pattern, compiling the number of stores patronised as well as the distribution of shopping activities. The main problem in distinguishing the two constructs lies in determining a clear cut-off point between patronage and loyal behaviour because both constructs are intertwined.

Yang and Peterson (2004) argued, that based on previous studies, it is extremely difficult to define and measure loyalty. Oliver (1999), stated that earlier conceptualizations of loyalty mainly focused on behavioural components (e.g. repeat purchases) and ignored the psychological elements.

Loyalty has been defined in various ways, but there is a wide acknowledgment for the approach of Jacoby and Chestnut (1978). The authors claimed loyalty is composed of three elements: 1) behavioural;

2) attitudinal; and 3) composite. The majority of the studies have focused on the behavioural component of loyalty. The behavioural component is perceived as the fundament of the concept of loyalty and mainly entails the degree of repetitive behaviour towards a same object (e.g. Neal 1999). The attitudinal element is considered as the consumers’ predisposition towards an object as a function of a psychological process (i.e. preference and commitment towards the object). An attitudinal loyal consumer is for example highly internally motivated to recommend the object to others (Sirohi et al.

1998). The composite component, states that neither the behavioural nor the attitudinal component by itself is sufficient to capture loyalty. In line with this conclusion, Laaksonen (1993) also posited that repurchase behaviour itself is not sufficient evidence of loyalty. Bloemer, de Ruyter, and Peeters (1998), also supported this statement and emphasised that consumers whose patronage is not based on loyalty, may exhibit attachment to the retailer attributes but easily switch to competitors in case they offer better products or services. Based on these fundamental ideas of loyalty, this research states that retail patronage intentions solely concentrate on the behavioural intention component of loyalty and do not

(8)

8 includes the psychological or attitudinal state at which the consumer feels connected with or committed to the product or service provider. Therefore, this research agrees with the earlier described statement of Davari et al., (2016), which stated that patronage intention predominantly captures the willingness to purchase at a particular store again and is considerably different from loyalty.

The decision has been made to focus on retail patronage intentions instead of retail loyalty intentions, for the main reason that this study focuses on intentions instead of actual measurable behaviour. Actual behaviour and the attitudinal component of loyalty are hard to measure at this stage of the technological development, due to the fact that no-checkout technology has not yet been widely available to consumers. As described, the attitudinal component of loyalty needs to be developed over time and could therefore be perceived as a complex mechanism and outcome of actual behaviour (e.g.

commitment towards the object) (Davari et al. 2016). In the context of this study, it is stated that at this stage, behavioural intentions can be captured (e.g. likelihood that a consumer will continue purchasing products from the store in the future, once no-checkout technology has been implemented), but it is less likely consumers can predetermine their attitudinal state when it comes to loyalty without experiencing the technology for a while (e.g. likelihood to recommend the grocery store or the increase of commitment towards the retailer).

Earlier identified antecedents of retail patronage intentions can be categorised into three groups:

1) product-relevant factors, which focus on product features and attributes; 2) market-relevant factors, which refer to characteristics of the retailer; and 3) personal factors, which relate to consumer characteristics (Pan and Zinkhan 2006). The main product-relevant factors influencing retail patronage intentions include, product quality (e.g. Darley and Lim 1993), price level (e.g. Dodds, Monroe, and Grewal 1991) and product assortment (e.g. Stassen, Mittelstaedt, and Mittelstaedt 1999). A considerable body of empirical research has been conducted on market relevant factors, crucial antecedents include, fast checkout (e.g. Thelen and Woodside 1997) and store image (e.g. Finn and Louviere 1996). Previous studies have also demonstrated that personal factors significantly affect retail patronage intentions, for example age (e.g. Roy 1994)) and store attitude (e.g. Eastlick and Liu 1997).

2.2 Utilitarian and Hedonic Shopping Values

Value has been a widely applied phenomenon in the area of consumer research. As a consequence, it has led to diverging interpretations of the concept. Dodds and Monroe (1985) stressed that the conceptualisation and interpretation of value varies across different contexts. In the early days, value had mainly been perceived as an outcome of a cost benefit trade-off and was widely applied in price- quality studies, estimating product choice (Babin, Darden, and Griffin 1994). Shopping was considered as a traditional product acquisition process, which was exclusively evaluated on the merit of the products and services purchased. This utilitarian type of shopping behaviour is perceived to be task-related, cognitive, rational and non-emotional. From a consumer perspective, utilitarian value mainly includes the evaluation of the accomplishment of a predetermined shopping need.

(9)

9 Bloch and Bruce (1984) were one of the first researchers emphasising the importance of intangible and emotional value during shopping. O’Guinn and Faber (1989, p.147) likewise identified an engagement oriented type of shopping behaviour, in addition to the traditional product acquisition view of shopping; “compulsive buyers buy not so much to obtain utility or serve from a purchased commodity as to achieve gratification through the buying process itself”. The authors linked compulsive buying to the desire, or need to use or experience a feeling or activity. This type of behaviour was for example connected to alcoholism, eating disorders and gambling. O’Guinn and Faber (1989) stated that this particular behaviour includes fantasies that enable an individual to escape from negative feelings.

Babin et al. (1994) extended the work of O’Guinn and Faber (1989) and placed these findings in a broader and more general context. The authors had revealed that hedonic values are “more subjective and personal than its utilitarian counterpart and resulting more from fun and playfulness than from task completion” (Babin et al. 1994, p.646). In the hedonic evaluation process, playfulness, arousal, fantasy, escapism and thus the experience are perceived to be far more important than task completion itself (Holbrook and Hirschman 1982). Following the research of Babin et al. (1994), it became widely accepted that the consumer shopping processes consist of both utilitarian and hedonic values (e.g. Jones and Reynolds 2006; Childers, Carr, Peck, and Carson 2001). However, the measurement scales developed by Babin et al. (1994) are generic and post-evaluation oriented. Due to the fact that no- checkout technology is currently not available to the general public, but value is perceived to be one of the most important antecedents of retail patronage intentions (Pan and Zinkhan 2006; Parasuraman and Grewal 2000), this study conducted research on the most important utilitarian and hedonic shopping value for no-checkout technology.

2.3 Convenience

As explained, utilitarian value is defined as the overall evaluation of benefits and sacrifices. In this assessment process, value for money (Zeithaml 1988) and convenience (e.g. Jarvenpaa and Todd 1997) are considered to be the most predominant cognitive determinants. As emphasised by McKinsey (2015), the main advantages of no-checkout technology concern time savings and convenience. However, it is of paramount importance that consumers perceive the technology as convenient in order to accept the technology and potentially positively affect behavioural intentions. In case it does not, it could jeopardise the intention to use the service and even stop shopping at the particular retailer.

The concept of convenience has been introduced by Copeland (1923) in the context of product categories. Convenience goods where perceived as products that require minimal time and physical and mental effort to purchase. Hence, convenience was used for describing the time and effort needed to purchase a particular product rather than an attribute of a product (Brown 1990). For decades, researchers have noticed that conserving time and effort is highly valued by consumers (e.g. Kelley 1958; Nickols and Fox 1983). Berry, Seiders, and Grewal (2002) stated that the increasing demand for convenience is inherent to socioeconomic changes, technological development, an increasingly

(10)

10 competitive business environment and opportunity costs that have risen with incomes. These findings have stimulated companies to develop more convenient products and services, and consequently encouraged consumers to use convenience as key dimension in purchase decision processes (Berry, Seiders, and Grewal 2002).

No-checkout technology can be perceived as a self-service technology (SST). Meuter, Ostrom, Roundtree, and Bitner (2000, p. 50) defined SST as “technological interfaces that enable customers to produce a service independent of direct service employee involvement”. Collier and Sherrell (2010, p.

492) had defined convenience in the context of SST as “the perceived time and effort required in finding and facilitating the use of a self-service technology.” In this context, the authors stated that convenience is the ability to decrease physical and sometimes cognitive effort to start a particular process, independent of employee involvement. Moreover, Collier and Sherrell (2010) and Ding, Hu, and Sheng, (2011) have shown that convenience is one of the driving factors of SST evaluation.

Canter (1983) developed the environmental psychology theory and claimed that the most important role of space, is its ability to facilitate the goals of its occupants. Based on this theory, Baker et al. (2002) demonstrated that store design has a great influence on the convenience perception of consumers. The authors demonstrated that a considerable large proportion of the shoppers’ goals relate to purchase the intended goods or services and getting in and out of the store as quickly as possible.

Empirical studies have already shown that convenience has a significant positive effect on customer satisfaction and behavioural intentions (Andaleeb and Basu 1994), customer switching behaviour (Keaveney 1995) and consumer perceptions and retention (Rust, Lemon, and Zeithaml 2004). Baker et al. (2002) provided evidence that the perception of the store design leads to lower convenience costs and eventually to higher retail patronage intentions. Seiders, Voss, Grewal, and Godfrey (2005) stated convenience is of great importance determining retail patronage behaviour. The authors provided evidence that convenience has a positive moderating effect on customer satisfaction and repurchase behaviour.

Convenience is a standalone construct that relates to a particular setting or in this case technology. Since no research has been conducted on the convenience perception of IoT retail technologies, but research towards comparable technologies (e.g. SST) offer sufficient grounds to state that convenience might positively affect retail patronage intentions, the following hypothesis is proposed:

H1: An increase of convenience perceptions, with regard to no-checkout technology, leads to an increase of retail patronage intentions.

2.4 Enjoyment

As described in section 2.2, shopping is not solely a utilitarian evaluation, instead there are various intangible and emotional elements affecting shopping processes. Baker, Levy, and Grewal (1992) stated

(11)

11 that shopping experiences contribute to affective reactions among customers. Baker et al., (2002) provided evidence that shopping experiences enhance retail patronage intentions. Concerning technological use, enjoyment is perceived as a key factor in the adoption process of a new technology (e.g Dabholkar and Bagozzi 2002; Childers et al. 2001). Weijters, Rangarajan, Falk, and Schillewaert (2007) stressed that fun is also an important component of enjoyment in the context of technology. The authors provided evidence that fun has a positive effect on the favourable attitudes towards SST.

Weijters et al. (2007, p.5) defined enjoyment as “the extent to which the activity of using technology is perceived to provide reinforcement in its own right, apart from any performance consequences that may be anticipated”.

Enjoyment in a shopping process is determined by consumers’ pleasure and excitement, triggered by store environments (Kim, Fiore, and Lee 2007). Swinyard (1993) provided evidence that a pleasant store environment and the quality of the shopping experience, enhances consumer engagement and increases shopping intentions. Various researches have demonstrated that consumers’ positive affective state does not only affect purchase behaviour (e.g. Babin and Darden 1996), but also other approach responses such as behavioural intentions (e.g. Donovan and Rossiter 1982). Multiple studies have provided evidence for a positive effect between affect and retail patronage intentions (e.g.

Donovan, Rossiter, Marcoolyn, and Nesdale 1994; Wakefield and Baker 1998). Hart, Farrell, Stachow, Reed, and Cadogan (2007) have shown that enjoyment of shopping experiences led to repatronage intentions in a shopping mall environment. Fiore, Jin, and Kim (2005) demonstrated that pleasure and arousal positively affect patronage behaviour in an e-commerce setting. Concerning online shopping, Chiu et al. (2009a) revealed that higher enjoyment perceptions result in higher purchase intentions.

However, Kim, Fiore, and Lee (2007) did not provide significant evidence for a positive relationship between enjoyment and patronage intentions in an e-commerce environment. Hence, discrepancies exist between evidence of various off- and online oriented studies. Because no research has been conducted on enjoyment perceptions with regard to IoT applications and the majority of the identified studies have revealed a positive relationship between enjoyment perceptions and behavioural intentions, the following hypothesis is proposed:

H2: An increase of enjoyment perceptions, with regard to no-checkout technology, leads to an increase of retail patronage intentions.

2.3 Privacy concerns

The concept of privacy has been deeply rooted within our society. Warren and Brandeis (1890) already described privacy as ‘the right to be alone’. However, a generic, agreed-on definition does not exist according to Phelps, Nowak and Ferrell (2000). The authors described the multidimensionality, as well as the various interpretations of the construct. Prosser (1960) introduced four dimensions of privacy: 1) intrusion; 2) disclosure; 3) false light; and 4) appropriation. This four-dimensional proposition of

(12)

12 privacy is still very relevant because the dimensions all relate to what consumers most often perceive as the main privacy issue, namely ‘information control’ (Phelps et al. 2000). For example, Krafft, Arden, and Verhoef (2017) stated that consumers indicate having serious concerns about which entities have access to their personal information, would like to be better informed about the use of data and are seeking for benefits in return for personal information disclosure. Smith, Milberg and Burke (1996) developed an approach that has specifically been geared to personal information privacy. The authors stated that privacy concerns of individuals relate to concerns about the: 1) collection; 2) improper access;

3) errors; and 4) secondary use of personal information. With regard to personal information and IT, privacy is defined in this study as; ‘the ability of the individual to control the terms under which personal information is acquired and used’ (Xu, Teo, Tan, and Agarwal 2009, p. 138).

Receiving a service in exchange for personal information has often been perceived and studied as a cost-benefit trade-off and is part of the social exchange theory (Homans 1961). The theory states that people only choose to engage in an exchange if they expect the net outcome to be positive. Since this theoretical view is extremely relevant in information exchange, the theory has often been applied in information privacy related studies (e.g Schumann, von Wangenheim, and Groene 2014). In these cost- benefit analyses, loss of privacy is obviously perceived as a cost. As a result, numerous studies identify the negative impact of privacy concerns on people’s willingness to share personal information (e.g., Son and Kim 2008). Likewise, Dunfee, Smith, and Ross (1999) stressed that the social contract theory provides a moral foundation for organisations. According to the authors, exchange of personal information could be perceived as a social contract, which is entered every time a consumer provides personal information to the organisation. Consequently, the contract is breached if the company collects or uses the information fraudulent. Tucker (2014) had for example shown that providing consumers with more control over data reduces the effects of privacy concerns. Furthermore, enhanced company reputation, trust and data protection seals are perceived to negatively affect privacy concerns (Xie, Teo, and Wan 2006).

Privacy concerns have been found to negatively affect consumer outcomes in various fields, such as mobile marketing and consumer loyalty (e.g. Demoulin and Zidda 2009; Zhao, Lu, and Gupta 2012; Krafft, Arden, and Verhoef 2017). In most cases, consumers perceive disclosure of private data as a personal sacrifice (Son and Kim 2008). Ackerman, Cranor, and Reagle (1999) conducted research on consumers’ attitudes towards online privacy and focused on how comfortable people were with revealing identity information to known or unknown organisations. The author found that this matter depends on the type of information and the perceived usefulness of the application. For example, Tucker (2014) found that a degree of personalisation increased privacy concerns (Tucker 2014). Chiu, Chang, Cheng and Fang (2009) provided evidence that low privacy concerns had a significant positive indirect effect on repurchase intention in the online context. Additionally, Chiu, Wang, Fang and Huang (2014) found that perceived risk (constructed by e.g. privacy concerns) negatively influenced buyers’ online repeat purchase intentions. However, Barkuus and Dey (2003) argued that numerous studies base their

(13)

13 study research on the common notion that a high degree of privacy is crucial for consumers, but the authors stress that few studies confirm this. For example, Forsythe and Shi (2003) found no significant effect between privacy concerns and patronage behaviour in an e-commerce context.

Nevertheless, Atzori, Iera, and Morabito (2010) stressed it is justified that consumers have serious IoT privacy concerns. The authors argued that the way data is collected and provisioned, is entirely different from the way it is done right now and the degree at which data is mined is even more extensive (e.g. compared to mobile applications). Atzori, Iera, and Morabito (2010) moreover stressed it will be impossible for consumers to fully control the personal information disclosure. It is impossible to account for these matters when sensor networks are used in IoT environments (Atzori, Iera, and Morabito 2010). For no-checkout technology this is the case, because cameras, RFID technology, sensors and smartphone are used to identify and track consumer behaviour. Hence, these statements are conflicting with the fundamental ideas of informational privacy concerns. Based on the findings of previous studies and serious privacy concerns relating to IoT applications, the following hypothesis has been constructed:

H3: An increase of privacy concerns, with regard to no-checkout technology, leads to a decrease of retail patronage intentions.

As explained, convenience perceptions relate to the desire for time and effort conservations and have a positive influence on purchase decision processes (Berry et al. 2002). It can be concluded that convenience, in the context of shopping behaviour, mainly relates to efficiency and effectiveness.

Privacy concerns are a psychological state that take into account the chances of conceivable negative outcomes. One could state that privacy concerns could jeopardise shopping efficiency and the intent to patronage due to high negative outcome perceptions. Therefore, high privacy concerns could lead to a more inefficient shopping process. Besides, cognitive effort is an important component of convenience (Collier and Sherrell 2010). In case high cognitive effort is required, patronage intentions could decrease.

Various studies conducted research on the moderating role of privacy concerns, predominantly in the context of online shopping. McCole, Ramsey, and Williams (2010) for example provided evidence that the relationship between trust and attitude towards online buying was moderated by the level of privacy concerns. Yun and Han (2013) have revealed that privacy concerns moderated the relationship between performance expectancy, which refers to the perceived usefulness of the technology, and continuous usage intention. The author however did not find a significant result for the moderating effect of privacy concerns on the relationship between effort expectancy and continuous usage intention.

Davari et al. (2016) demonstrated that convenience increased online retail patronage trough the mediating variable ‘service quality’. The authors found that the positive relationship between service quality and patronage intentions was negatively moderated by security concerns (e.g. privacy concerns).

(14)

14 Moreover, Chiu et al. (2014) found a negative moderating effect of perceived risk on the relationship between utilitarian shopping value and repeat purchase intentions. Based on these findings and the statement that privacy concerns make the shopping process more inefficient, the following hypothesis is proposed:

H4: Privacy concerns moderate the relationship between convenience and retail patronage intentions such that the relationship is less strong for consumers with high privacy concerns than for consumers with low privacy concerns.

As explained in the section above, a moderate amount of research has been conducted on the moderating role of privacy concerns in behavioural studies. However, these studies predominately focused on utilitarian oriented relationships, less research has been conducted on hedonic oriented topics. Chiu et al. (2014) hypothesised that perceived risk positively moderated the relationship between hedonic value and online repeat purchase intentions. The authors justified the relationship by comparing the process to gambling or purchasing lottery tickets. According to the authors, consumers feel entertained by risking monetary losses when experiencing high arousal during uncertainty and winning. Nonetheless, the authors did not find a significant result. Though, monetary losses are subsequently different from privacy losses. From a consumer perception, it is easier to understand what the consequences of potential monetary losses in an e-commerce setting are compared to privacy losses in an IoT setting.

In the context of technological use, enjoyment concentrates on providing reinforcement apart from any performance consequences (Weijters et al. 2007). In case consumers perceive the technology to be enjoyable it is stated that this will enhance retail patronage intentions. However, this relationship could be influenced by possible losses. Previous studies have shown that in general, individuals who harbour strong concerns about specific issues require compelling arguments to modify their belief structure (Boritz, No, and Sundarraj 2008; Lau, Smith, and Fiske 1991). The stronger the concerns, the a more persuasive message with strong evidence is required in order to overcome the associated apprehension (Angst and Agarwal 2009). Enjoyment and hedonic shopping values in general are not perceived as strong evidence and its effects are therefore vulnerable to be influenced by strong privacy concerns. Following the prospect theory (Kahneman and Tversky 1979), which explains that risks interact with the value of potential gains in predicting behaviour, it states that losing hurts more than gaining. Abdellaoui, Bleichrodt, and Paraschiv (2007) provided evidence that from a psychological perspective, losses are perceived as twice as powerful as gains. In general, risk perceptions can affect individuals’ feelings (Yüksel and Yüksel 2007). Mainly as a result of the negative consequences associated with it. As a result, negative feelings can cause anxiety, discomfort and uncertainty (Featherman 2001; Dowling and Staelin 1994). In the context of shopping, higher risk perceptions negatively affect pleasure, which concern the degree to which a consumer feels good, joyful and satisfied in the situation (Yüksel and Yüksel 2007). It is expected that high privacy concerns about no-checkout

(15)

15 technology outweigh the potential enjoyment gains and therefore negatively influence the relationship between enjoyment and retail patronage intentions:

H5: Privacy concerns moderate the relationship between enjoyment and retail patronage intentions such that the relationship is less strong for consumers with high privacy concerns than for consumers with low privacy concerns.

The literature review and hypotheses have led to the following conceptual model:

3. Methodology

3.1 Data collection

Secondary data was not available for testing the hypotheses of this study, consequently primary data has been gathered. An online survey has been used for data collection and was preferred over a paper survey due to flexibility, accuracy and efficiency reasons (Lumsden and Morgan 2005). The online survey was hosted via ‘www.qualtrics.com’. Qualtrics has been selected as most appropriate software solution since it has longstanding experience with facilitating surveys, the University of Groningen has a partnership with Qualtrics and various reliable multimedia functionalities are offered, which are required for the purpose of this study. All of these factors minimise the chance for an instrument bias (Blumberg, Cooper, and Schindler 2011).

The survey has predominantly been spread among relatives of the researcher. Due to this arbitrary and subjective procedure, the sampling method has been characterised as nonprobability sampling (Blumberg et al. 2011). More specific, this research has used both convenience and snowball sampling. Convenience sampling has been applied due to the fact that respondents were contacted because of their relative ease of access (e.g. friends, colleagues and family). Moreover, snowball sampling has been used since a few participants were kindly asked to share the link of the survey with acquaintances (Blumberg et al. 2011). Participants were invited to participate in the study via e-mail,

Figure 1. Conceptual model

(16)

16 social media or Whatsapp message. However, data collection via these types of tools have been criticised for the non-response bias (Armstrong and Overton 1977). The non-response bias could jeopardise the external validity of the study. Armstrong and Overton (1977) demonstrated that among the people that have been invited to participate in a survey, results of people that participate and did not participate significantly could differ due to personal characteristics and behaviour (Armstrong and Overton 1977).

Therefore, it is important to test for the non-response bias because it increases the generalisability of the results (Blumberg et al. 2011). In order to minimise the non-response bias, people who have received a survey invitation via convenience sampling also received a follow-up message with a kind reminder to fill out the survey. Armstrong and Overton (1977) provided evidence that participants who fill out a survey at a later moment in time are more similar to non-respondents compared to particpants who fill out the survey shortly after the survey has been distributed. Respondents that participated at a later moment in time often needed an additional stimulus, such as a reminder, to fill out the survey. The sample of this study has chronologically been diveded into five segments in order to test for the non- response bias (Armstrong and Overton 1977). An independent samples t-Test has been performed in order to determine whether the mean score of the dependent variable of the first segment, significantly differs from the mean score of the fifth segment. To legitmatise the test, the reseacher aimed to distribute the survey to all particpants within the first three days of the data collection process so the sample could be perceived as a homogenous group. As shown in Table 2 of Appendix A, no significant difference exists between the mean score of the first and fifith segment (p > .1). Therefore, it can be concluded that no severe non-repsonse bias exists.

Respondents were tested for two participation restrictions: 1) the participant should be Dutch and 2) the participant must be at least 18 years old. Non-Dutch citizens were excluded from participation, since various studies have provided evidence that cultural differences exists in relationships between privacy concerns and behavioural intentions (e.g. Lowry, Cao, and Everard 2011). A minimum age of 18 was a prerequisite for participation because retail patronage intentions concern purchase and financial behaviour. Minors are excluded from participation since they often financially depend on their parents’

when it comes to grocery shopping. Moreover, from an ethical perspective they will probably not be allowed to create an account for no-checkout technology as it requires a direct connection with a credit account.

Because this research concerns a technology that is not commonly known among consumers, a short movie about no-checkout technology has been provided prior to filling out the questionnaire. Since video incorporation increases the chance for an instrument bias, the information supplied in the video was provided in textual form after the video as well. Thus, in case participants have watched the video in a noisy environment they would still be fully informed about the technology and able to proceed with the survey. The video was subtitled and the meaning of potentially ambiguous concepts were clarified.

After the video and textual summary about no-checkout technology, a clear scenario was provided. This scenario instructed participants to keep two important facts in mind, while answering the questions of

(17)

17 the survey: 1) the no-checkout technology will be implemented at the supermarket that the participant visits most frequently and 2) the no-checkout technology will replace the ordinary payment option to pay via a cashier at a cash desk. It has been decided to choose for a scenario at which the technology will be implemented at the most frequently visited supermarket. Internal validity could be jeopardised in case a scenario was chosen in which the technology would be implemented at a competitor of the

‘most frequently visited supermarket’ or a completely new supermarket (e.g. Amazon Go). External factors such as loyalty characteristics (e.g. strong disgust towards the retailer) might in this case have great influence on the relationships of the conceptual model and these are not part of the scope of this research. Apart from that, the decision has been made to state that no-checkout technology will fully replace the option to pay via a cashier at a cash desk because this will eventually be the case should the technology become successful. Furthermore, this also forced participants to truly concentrate on their benefit-and-loss perception of the technology, which is not influenced by for example the following thought: ‘sure, I like this technology as long as I can just pay at the counter’.

The survey has been translated into Dutch and was checked by a professor and two MSc students of the University of Groningen. Furthermore, backward translation has been used, assuring linguistic validity. The survey has been provided in Appendix B. All questions of the constructs have been randomised in order to enhance construct validity. Furthermore, some of the original measurement instruments used for this study contained reverse-scored questions (e.g. one negatively worded question compared to multiple positively worded questions). The decision has been made to include these reverse-scored questions in the survey because it minimises the response bias (Blumberg et al. 2011).

The study has been conducted on individual level. According to Hair, Black, Babin, and Anderson (2010) a minimum amount of 5 respondents per item (question) of the conceptual model was required, resulting in a minimum of 85 respondents. As will be explained in more detail in section 3.3, the total sample consisted of 138 respondents.

3.2 Measurement Instruments

The measurement instruments of this study have been selected on the basis of: 1) the reputation of the journal in which the measurement scale was published; 2) the frequency of use by other researchers;

and 3) applicability of the instrument in the context of this study. As a result of extensive research, the following measurement instruments have been selected for the constructs of this study.

Retail Patronage Intention – As explained in the literature review, patronage intentions have been measured with a variety of instruments. Metrics that have been used to conduct research on patronage intentions are similarly used to measure for example loyalty intentions. Various studies (e.g.

Chiu et al. 2009b) measured loyalty intentions, with an instrument that solely covered the behavioural component of loyalty and therefore actually measured patronage intentions. Patronage and loyalty oriented measurement instruments have been ambiguously used and therefore measurement instruments of both constructs have been included and evaluated in the search for the most appropriate metric for

(18)

18 this study.

After extensive research and application of the predetermined instrument criteria, two measurement scales were perceived to be appropriate for measuring retail patronage intention in the context of this study. Firstly, the measurement scale developed by Dodds, Monroe and Grewal (1991), and has for example been used by Baker et al. (2002) and Grewal et al. (2003) in order to measure store patronage intentions. Secondly, the metric developed by Pavlou (2003), which was also frequently used in various studies for example by Chang and Chen (2008) to measure purchase intention and Chiu et al.

(2009b) to measure loyalty intentions. It has been decided to use the measurement instrument of Pavlou (2003) as the instrument of Dodds, Monroe and Grewal (1991) tends more towards measuring elements of loyalty (e.g. item: ‘I would be willing to recommend this store to my friends’) compared to the measurement scale of Pavlou (2003), which solely focuses on behavioural intentions. Moreover, the measurement scale of Pavlou (2003) has previously been used in studies that have a similar scenario and context as this research. For example, Chiu et al. (2009b) used the instrument of Pavlou (2003) for stores that are already familiar to the participant (e.g. item: ‘I will continue purchasing products from the online store in the future) compared to measurement scale of Dodds, Monroe, and Grewal (1991) which has mainly been used for concepts or stores that have not yet been patronised by the participant (e.g. item ‘the likelihood that I would shop in this store is very high’). Chiu et al. (2009b) adopted version of the scale of Pavlou (2003) has been used for this study instead of the original instrument of Pavlou (2003), due to the fact that less modifications were required.

The measurement scale consists of three items and has been set out on a 1-7 Likert-scale; where 1 indicates strongly disagree, 4 indicates neutral, and 7 indicates strongly agree: e.g. item ‘When the no- checkout technology has been implemented, I intend to continue purchasing products from the store in the future’. An overview of all items of the instrument is provided in Table 2 of Appendix C.

Convenience - The construct of convenience has been used in various fields. In the context of technology, the measurement scale of Childers et al. (2001) and Mathwick, Malhotra, and Rigdon (2001) has most frequently been used. For example, Chiu et al. (2014) conducted research on the online shopping perception, using the convenience measurement scale of Childers et al., (2001). The measurement scale of Mathwick, Malhotra, and Rigdon (2001) was for example used by Kleijnen, de Ruyter, and Wetzels (2007) and Tojib and Tsarenko (2012) in order to measure time convenience perceptions in the context of smartphone use. The metric of Mathwick, Malhotra, and Rigdon (2001) has initially been designed as an efficiency scale and has therefore most often been used for measuring time convenience. Since the measurement scale of Childers et al. (2001) covers a broader perspective of the convenience construct, by not solely focusing on time (e.g. ‘using TAS would be a convenient way to shop’), it is perceived to be more appropriate for the purpose of this study.

The construct consists of four items and has slightly been adopted to the technological context of this study. The items were measured on a 1-7 measurement scale; where 1 indicates strongly disagree, 4 indicates neutral, and 7 indicates strongly agree: e.g. item ‘No-checkout technology would allow me

(19)

19 to save time when shopping’. All items of the constructs can be viewed in Table 3 of Appendix C.

Enjoyment –The three item enjoyment scale of Davis, Bagozzi, and Warshaw (1992) and four item scale of Dabholkar (1996) have most frequently been used as enjoyment metric in the context of technology (e.g. Demoulin and Djelassi 2016; Hwang and Kim 2007; Venkatesh 2000). The instrument of Davis, Bagozzi, and Warshaw (1992) has been developed in order to measure the effect of perceived enjoyment on usage intention. The metric of Dabholkar (1996) has been designed in the context of SST and quality perceptions. Both instruments have incorporated a general enjoyment and fun component (e.g.’I find using the system to be enjoyable’). However, the metric of Dabholkar (1996) has been selected as most suitable measurement scale for this study because the metric includes the descriptors

‘entertaining’ and ‘interesting’. The author posited that both descriptors enrich the instrument when it is used in innovative settings, as these elements capture the novelty aspect of enjoyment. Moreover, the measurement instrument of Dabholkar (1996) is pre-evaluation orientated (e.g. ‘using a touch screen for self-service will be entertaining’) compared to the post-evaluation oriented metric of Davis, Bagozzi, and Warshaw (1992) (e.g. ‘I have fun using the system’). Due to the fact that participants have no physical experience with no-checkout technology, the metric of Dabholkar (1996) required minor modifications in order to be suitable for the purpose of this study.

The enjoyment construct consists of 4 items. The items were measured on a 1-7 measurement scale; where 1 indicates strongly disagree, 4 indicates neutral, and 7 indicates strongly agree: e.g. item

‘using no-checkout stores will not be fun’. All items of the enjoyment measurement instrument have been presented in Table 4 of Appendix C.

Privacy Concerns – Privacy concerns studies have often focused on specific research areas or technological applications. As a result, measurement instruments have often been considerably modified or even designed for a specific study (e.g. Chellappa and Sin 2005). Privacy concern instruments are in most cases quite comprehensive, as they are frequently constructed by detailed sub-dimensions and therefore become inapplicable (e.g. Stewart and Segars 2002). A generic instrument has been considered as more appropriate for the purpose of this study. The measurement instrument of Dinev and Hart (2004) has been widely applied (e.g. Fogel and Nehmad 2009), as well as the concern for information privacy scale (CFIP) of Smith, Milberg, and Burke (1996) (e.g. Xu and Gupta 2009). The CFIP measurement instrument of Smith, Milberg, and Burke (1996) has been used for this study. Compared to the scale of Dinev and Hart (2004), less modifications were required to adopt the scale to the purpose of this study (e.g. item of Dinev and Hart (2004): ‘When I shop online, I am concerned that the credit card information can be stolen while being transferred on the Internet’).

The measurement scale of Smith, Milberg, and Burke (1996) consists of five items and has been slightly modified by Xu (2007). This modified version was for example also used by Yun and Han (2013). This research has used the modified version of Xu (2007) because the context of that study is comparable to this one. The metric has been adopted from a location based technology towards a no- checkout technology oriented scale. The five items are measured on a 7-point Likert-scale; where 1

(20)

20 indicates strongly disagree, 4 indicates neutral, and 7 indicates strongly agree: e.g. ‘I am concerned that the retailer may keep my personal information in a non-accurate manner in their database’. All items and modifications are provided in Table 5 of Appendix C.

Control variables – To control for the effects of possible confounding variables and improve the internal validity of this study, multiple control variables were measured and analysed. Three demographic characteristics (age, gender and level of education), two personal traits (risk propensity and personal innovativeness in IT), retailer commitment towards the selected retailer and the need for interaction were incorporated as control variable due to recommendations of previous studies and the scenario of the study design.

The demographic factors, age, gender and level of education have been used as control variables by numerous patronage oriented studies (e.g. Nesset, Nervik, and Helgesen 2011; Yang and Peterson 2004). Risk propensity and Personal innovativeness in IT are personal traits and therefore perceived to be stable over time, across situations and are not influenced by external factors (Agarwal and Prasa 1998; Cho and Lee 2006). Risk propensity refers to an individual’s general tendency towards avoiding and taking risks (Chang and Chen 2008). Risk-prone consumers enjoy overestimating the probability of losses compared to potential gains, which influences behaviour intentions (Chang and Chen 2008; Xu, Teo, and Tan 2005). Because privacy concerns relate to potential losses and convenience and enjoyment to potential gains, it is expected that risk propensity might influence retail patronage intention and has therefore been incorporated as control variable. The measurement instrument of Donthu and Gilliland (1996) has been selected for this study since, compared to other instruments, minor modifications are required to serve the purpose of this study (e.g. Meertens and Lion 2008). Personal innovativeness in IT refers to ‘the willingness of an individual to try out any new information technology’ (Agarwal and Prasa 1998. p. 206). Xu, Teo, and Tan (2005) posited that personal innovativeness in IT has a significant influence on behavioural intentions in case the research concerns a new type of technology and should therefore be controlled for in the analyses. The measurement scale of Agarwal and Prasad (1998) has been selected as most suitable instrument for this research because other personal innovativeness instruments do not solely focus on IT (e.g. Flynn and Goldsmith 1993) and it has frequently been used by various studies (e.g. Sun 2012). As this study incorporates a scenario in which the consumer is forced to select the retailer he/she most frequently visits, the analyses have been controlled for the commitment one has towards the selected retailer. This attitudinal loyalty characteristic could significantly influence behavioural intentions. The measurement scale of Ray and Chiagouris (2009) has been selected due to the fact that the attitudinal component of loyalty (commitment) clearly stands out from the behavioural components. Lastly, the need for interaction with the cashier has been incorporated as control variable because the scenario states that technology replaces the option to pay via a cashier at a cash desk. The measurement instrument of Collier and Sherrell (2010) has been perceived as most suitable instrument for this study as it is a slightly modified version of the frequently used need for interaction instrument of Dabholkar (1996). The instrument of Collier and Sherrell (2010) specifically focused on checkout

(21)

21 desks and SST. As a result, only minor adjustments were required to serve the purpose of this study. All control variables were tested for both the direct effects, as well as the moderating effects, on retail patronage intentions. The items have been set out on a 7-point Likert-scale; 1 indicates strongly disagree, 4 indicates neutral, and 7 indicates strongly agree. An overview of the instruments of the control variables is provided in Table 6-9 of Appendix C.

3.3 Analyses

Data preparation has been executed before the start of the analyses. As a first step, the dataset of 173 respondents was imported from Qualtrics into IBM SPSS Statistics 24 (hereinafter referred to as SPSS) and checked for incomplete surveys. Accordingly, 25 surveys have been identified as incomplete and were deleted. Moreover, 10 participants have indicated that they were not able to view the video and have therefore also been excluded from the dataset. None of the participants were younger than 18, which led to a final dataset of 138 respondents (n=138). As next step, reverse-scored questions (PIIT_3, NFI_3, ENJ_1 and ENJ_3) were reverse-coded. All analyses of this study have been performed with SPSS due to the fact that the relationships of the conceptual model can be tested with analyses offered in SPSS, the program is considered reliable and the researcher of this study has prior experience with the software.

The overall aim of the performed analyses was to determine whether the proposed hypotheses of this study can be accepted or rejected. A multiple regression analysis has been considered the most suitable analysis to serve this purpose. However, before performing the multiple regression analysis, additional analyses have been executed in order to enhance reliability and minimise construct validity.

Firstly, sample statistics have been presented in order to provide insight into the general characteristics of the sample. Secondly, a reliability analysis has been performed in order to define the internal consistency of the items of each construct. Hence, the analysis determined whether the proposed items measured the same construct and if the measurement scale could therefore be considered reliable (Blumberg et al. 2011). A Cronbach’s Alpha cut-off point of >0.7 has been applied for this assessment (Blumberg et al. 2011). Thirdly, an Exploratory Factor Analysis (EFA) has been executed. Before the start of the EFA, a Kaiser-Meyer-Olkin Measure of Sampling Adequacy (KMO) and Barlett’s test of sphericity was performed in order to determine whether the data of this study was suitable for a factor analysis. A KMO value of >0.5 and significance level <.05 at Barlett’s test of sphericity were required in order to perform the EFA. The EFA provides a factor structure by determining the correlation between the items of the dataset. The analysis determines which items explain most of the variance of each construct. The conceptual model of this study contains four constructs, therefore an EFA with a fixed amount of four factors has been performed. Any other number of factors is not supported by the literature and would therefore not be useful for further analyses. The EFA determined that certain items had to be deleted from the measurement instruments due to cross loading and applying a coefficient’s cut-off point of <0.4 (Hair, et al. 2010). Furthermore, the EFA determined whether the four constructs are perceived

(22)

22 as unrelated and hence no severe discriminant validity exists. Fourthly, after deleting the items, a second reliability analysis has been performed in order to ensure that all constructs were perceived reliable.

Fifthly, descriptive statistics have been presented in order to provide insight into the general statistics of the constructs. Sixthly, a Pearson product-moment correlation analysis has been performed in order to test for the strength of the linear association between each of the newly created variables (Blumberg et al. 2011). High correlations might have indicated the existence of multicollinearity. Together with the Variance Inflation Factor (VIF) of the multiple regression analysis, it is determined whether severe multicollinearity exists among the identified high significant correlations. Finally, the multiple regression analysis has been carried out in order to determine whether evidence has been provided for a significant linear relationship for each of the proposed hypotheses. Hence, the analysis for example provides evidence whether consumers with higher privacy concerns, with regard to no-checkout technology, have lower retail patronage intentions. To ensure that the model improved overall explanatory power, three models have been compared. As shown in Table 3 of Appendix D or Table 4, the baseline model only consists of the control variables and independent variable (Model 1). Model 2 adds the main effects of the independent variable. Model 3 covers the elements of the prior models and the moderating effects. As recommended by Aiken and West (1994) and Kenny and Judd (1984), the independent variables of this study have been mean centred for measuring the moderating effects. As part of the multiple regression analysis and in order to assure the reliability of this study, the model has been tested for multicollinearity. A VIF cut-off point of <10 has been maintained (Hair, et al. 2010). As shown in Table 3 of Appendix D, the VIF score of Model 3 varies from 1.116 and 1.502. Therefore, it can be concluded that no high inter-correlations or inter-associations exist among the independent and control variables and thus no severe multicollinearity exists.

4. Results

4.1 Sample Statistics

The sample (n=138) was composed of 58.7% male and 41.3% female respondents. The average age of the participants was 32.63 years old, however 64.2% of the respondents was 26 years or younger.

Concerning the level of education, 5.8% percent of the respondents have indicated to have a secondary school and/or intermediate vocational education (MBO) degree and 94.2% a polytechnic (HBO) and/or University (WO) degree. None of the respondents has indicated to have primary education as highest level of education. Besides, participants were asked which retailer they most frequently visit. 76.1% of the respondents most often shop at Albert Heijn, 10.6% of the participants at Plus and 7.2% at Jumbo.

4.2 Reliability Analyses, Exploratory Factor Analysis, Descriptive statistics, and Correlations As explained in section 3.3, a reliability analysis has been performed in order to determine the reliability of the measurement scale, constituted by the items of each instrument. As shown in Table 1 and for the

Table of Contents

Table of Contents