Would you trust this webshop? An empirical analysis of consumer trust antecedents in mobile commerce websites in the Netherlands

(1)

U

NIVERSITY OF

A

MSTERDAM

M

ASTER

T

HESIS

Would you trust this webshop? An

empirical analysis of consumer trust

antecedents in mobile commerce websites

in the Netherlands

Author:

Paul VERWEY

Supervisor: Dr. Dick HEINHUIS

A thesis submitted in fulfilment of the requirements for the degree of Master of Science

in the

Information Studies Business Information Systems

(2)

i

University of Amsterdam

Abstract

Faculty of Science Business Information Systems

Master of Science

Would you trust this webshop? An empirical analysis of consumer trust antecedents in mobile commerce websites in the Netherlands

by Paul VERWEY

In this study, customer trust in a mobile commerce website was investigated from the literature. More specifically, which website factors influence customers’ trust during online shopping. Literature on the subject has shown that there are sev-eral trust antecedents in mobile commerce websites, namely usefulness, ease of use, customization, security, design and content. While these constructs have shown to be determinants of trust in several countries, they were not yet researched in the Netherlands. Therefore, the research question of this thesis was "What are the factors that affect customer trust in a mobile commerce website in the Netherlands?". Based on existing literature, a model was developed that incorporated these trust antecedents regarding a website. The results show that usefulness, ease of use, customization, design and content were all significant antecedents to trust in a mobile commerce website. Finally, no support was found for the hypothesis that security is an an-tecedent to trust in a mobile commerce website.

(3)

ii

Acknowledgements

I would like to sincerely extent my gratitude to Dick Heinhuis for giving me feed-back throughout the creation of my master thesis. I would also like to thank every-one else who has helped me every-one way or another.

Lastly, I would like to thank team 3 a.k.a. team hondje (you know who you are) for all the great times and pushing me during the hard times. Without you guys, this thesis would not exist.

(4)

1

Chapter 1

Introduction

Electronic commerce (e-commerce) has revolutionized industries since it first gained prevalence in the 1990’s (Chong, Chan, and Ooi, 2012). Examples are Amazon and eBay, who have dramatically altered the way we buy things. With the growth of wireless communication technologies, mobile commerce (m-commerce) is now seen as the new business model and platform that will have a similar, if not bigger, im-pact on industries than e-commerce (Chong, Chan, and Ooi, 2012). M-commerce is seen as an extension of e-commerce, whereby the transaction of business is con-ducted using mobile devices, such as smartphones (Wong and Hsu, 2008; Ngai and Gunasekaran, 2007). This provides consumers with the advantage of having access to information anywhere, any time, any place.

Most people today cannot imagine a day without their smartphone. The device is used during the day for reading email, sending messages, shopping online and social media among other things. So much so, that it is becoming the main device of choice for many online activities (Ofcom, 2016). Where the desktop or laptop used to be the main device of use, these days on average users spent more than twice as much time on a smartphone than on their laptop or desktop. In 2016, the average time spent browsing on a laptop or desktop is 28 hours per month, where it is 65 hours on a smartphone (Ofcom, 2016)1_.

However, even though mobile usage continues to grow, and is recognized by retailers to contribute to an average of 50% of in-store revenue, mobile purchases continues to lag (McGuire, 2016). The Gartner report states: “a massive increase in mobile traffic, with more than 50% year-over-year growth from 1Q14 to 1Q15, has failed to bring a commensurate increase in sales from mobile channels". Although mobile accounted for 60% of total time spent shopping online in the U.S., 87% of online retail purchases were made on desktops (McGuire, 2016). In the Netherlands in Q3 2016 only 5% of the online purchases were made on a smartphone, compared to 80% on desktop (Thuij, 2016). It can therefore be argued that the purchase intent is lower on smartphones than on desktops.

A factor that has shown to be an important predictor of purchase intent is trust (Gefen, Karahanna, and Straub, 2003; Nilashi et al., 2015). Trust is an essential com-ponent in any relationship: interpersonal, social structures as well as in business relationships. This is especially the case in interactions where uncertainty and de-pendency is involved, such as purchasing products or services. The acceptance of online transactions on mobile devices depend on people’s trust in the technology used for the transactions (e.g. the device and mobile network) and in organizations as the other parties in the transactions. Transactions like these are characterized as faceless and intangible, which could result in people’s reluctance to engage in any form of online transaction (Beldad, De Jong, and Steehouder, 2010). It could be that

1_{Average of eight (developed) countries. Statistics of the Netherlands specifically could not be}

(5)

Chapter 1. Introduction 2

the slow increase in sales from mobile channels is a result of a lack of trust in m-commerce.

Past m-commerce research has looked at the adoption and success factors by utilising e.g. the Technology Acceptance Model (TAM) of Davis (1989). These in-clude e.g. Cyr, Head, and Ivanov (2006), Feng, Hoegler, and Stucky (2006), Wong and Hsu (2008), and Wu and Wang (2005). This could have to do with m-commerce being a new innovation back then and therefore acceptance factors and perceived usefulness of the technology itself was an important topic. This was also in a time where phones had low resolutions, small screens, tiny keyboards and where the mobile networks were limited in bandwidth, connection stability and function pre-dictability (Siau and Shen, 2003; Vance, Elie-Dit-Cosaque, and Straub, 2008; Wu and Wang, 2005). Since then, devices and mobile networks have come a long way. Mo-bile phones have bigger screens, higher resolutions and moMo-bile networks no longer have limited bandwidth and are now more stable than ever. It could be that because the technology itself (the device and the network) matured and users have broadly adopted it, it is no longer regarded as a trust issue. This thought is also recognized by Siau and Shen (2003), who state: "as mobile technology evolves, focus will shift from engendering customer trust in technology to engendering trust in vendors".

1.1 Problem statement

For online organizations, the website is the primary means by which consumers de-cide whether or not to conduct business with that vendor (Pengnate and Sarathy, 2017). Although a website on a mobile device is still a website, it does provide unique features over websites viewed on traditional desktops. Differences are e.g. that the interaction style is unique due to the constraints of the device (such as screen size) and the usage patterns also differ from those of traditional desktop computers (Feng, Hoegler, and Stucky, 2006). According to Junglas and Watson (2006), m-commerce provides five features that makes it unique from e-commerce, namely portability, reachability, accessibility, localisation and identification (Junglas and Watson, 2006).

While trust in e-commerce websites is a much researched area (e.g. McKnight, Choudhury, and Kacmarc (2002) and Beldad, De Jong, and Steehouder (2010)), lit-erature specifically focused on m-commerce websites is limited. Only two recent research papers published in a top journal in the context of m-commerce trust have been found. Li and Yeh (2010) have found that, in Taiwan, design aesthetics sig-nificantly impact website characteristics components which sigsig-nificantly affect cus-tomer trust in mobile websites. Nilashi et al. (2015) found a strong influence of se-curity, design, and content in mobile commerce websites on trust, but they also state that the proposed model can be improved to consider more factors through literature study and survey. This research was carried out in Malaysia.

Research, however, shows that these trust antecedents are not culturally neutral. According to Vance, Elie-Dit-Cosaque, and Straub (2008), multiple cross-cultural studies have shown that cultures have different preferences with regard to the de-sign of an IT artefact and aspects of website dede-sign. Aspects such as navigability, layout, and graphical elements were preferred differently across Japanese, Cana-dian, U.S., and German cultures (Vance, Elie-Dit-Cosaque, and Straub, 2008; Cyr, 2013). Vance, Elie-Dit-Cosaque, and Straub (2008) state that this has important im-plications, because this shows that "individuals of different cultures may exhibit markedly different attitudes toward placing trust in an IT artefact, which may, in

(6)

Chapter 1. Introduction 3

turn, translate into varying levels of intention to adopt the IT artefact" (Vance, Elie-Dit-Cosaque, and Straub, 2008). These findings suggest that trust antecedents that were found significant for Malaysian and Taiwanese consumers, may not necessarily translate or apply to the Dutch consumers, whom are the subject of this thesis.

A culture aspect that has been shown to be closely related to trust, is uncertainty avoidance (Vance, Elie-Dit-Cosaque, and Straub, 2008). For instance, Cyr (2013) has used uncertainty avoidance for its relevance to website design, as well as to trust and security issues related to e-commerce. Uncertainty avoidance is the degree of which members of a society feel uncomfortable with uncertainty and ambiguity (Hofst-ede, 2010). A difference in uncertainty avoidance scores can be noticed between the Netherlands and the cultures researched in the aforementioned papers on m-commerce trust (Li and Yeh (2010) and Nilashi et al. (2015)). The Netherlands score 53, Taiwan 69 and Malaysia 36. These scores mean that the Netherlands have a slight preference for avoiding uncertainty, Taiwan a high preference and Malaysia a low preference for avoiding uncertainty (Hofstede, 2010).

Considering these findings, more research on the subject of mobile commerce websites and which factors affect consumer trust in the Netherlands could therefore add knowledge to the current state of the literature.

1.2 Research question

This research aims to improve understanding of online trust antecedents in the con-text of m-commerce in the Netherlands. Therefore, the following research question is answered in this thesis:

What are the factors that affect customer trust in a mobile commerce websites in the Nether-lands?

To answer this question, the following sub-questions need answering:

1. Which website factors play a role in building consumer trust in mobile com-merce websites?

2. How are the website factors related to trust?

3. How can the proposed conceptual model be empirically verified?

1.3 Practical relevance

The usage of smartphones has increased rapidly in the last decade and m-commerce is playing a more and more important role. The understanding of the different trust antecedents in this context, however, is limited. The topic of this thesis is, therefore, relevant for online merchants that face challenges with the trustworthiness of their website and purchase intent of customers on smartphones. It aims to provide new perspectives that online merchants can use to provide a more trustworthy experi-ence to their customers, and hopefully with that, increase their sales.

(7)

4

Chapter 2

State of Knowledge

In order to answer the first sub-question, the concepts mobile commerce and trust need to be examined. The following is a brief description of what mobile commerce entails and which definitions are used throughout the literature. A more detailed account of online trust and a definition is given in the succeeding section.

2.1 Mobile commerce

Mobile commerce (m-commerce) is a relatively new phenomenon. The following goes into the various definitions the literature has used to define m-commerce.

2.1.1 Definitions of mobile commerce

Researchers have adopted a broad definition of m-commerce to explore its potential benefits. Often times m-commerce is understood as mobile e-commerce with the use of wireless technology, particularly mobile devices and mobile internet, to facilitate activities such as (business) transactions, information search and communication (Feng, Hoegler, and Stucky, 2006). For example, Chong, Chan, and Ooi (2012) define m-commerce as “any transaction, involving the transfer of ownership or rights to use goods and services, which is initiated and/or completed by using mobiles access to computer-mediated networks with the help of mobile devices”. In other defini-tions, m-commerce is seen as an extension of e-commerce whereby transactions are done on wireless instead of wired networks, and from fixed locations to any time, anywhere, and any device (Feng, Hoegler, and Stucky, 2006). Others leave out the device part and refer to "any transaction with monetary value that is conducted via a mobile network" (Ngai and Gunasekaran, 2007). In this definition, a desktop that is connected to a mobile network would also be considered m-commerce. Feng, Hoe-gler, and Stucky (2006) also include in their definition that the business transactions are data-driven: "Mobile commerce refers to all data-driven business transactions and exchanges of wireless via by users of mobile devices value telecommunication networks". Based on these definitions, it can be concluded that common elements in the m-commerce definitions are: transactions, wireless networks, mobile devices and exchange of goods and services. In most definitions there is consensus that the medium used to conduct the transactions are wireless networks and mobile devices (such as (smart)phones and more recently, tablets).

The definition of m-commerce considered appropriate for this thesis follows that of Chong, Chan, and Ooi (2012), with the exception that the mobile device in this thesis is considered a smartphone. The m-commerce definition used is therefore: any transaction, involving the transfer of ownership or rights to use goods and services, which is initiated and/or completed by using mobiles access to computer-mediated networks with the help of a smartphone.

(8)

Chapter 2. State of Knowledge 5

2.2 Trust

The following is a brief description of the definition of trust that is used in this thesis and, since trust is a broad concept, which type of trust this thesis focusses on and how it can be achieved.

2.2.1 Definition of trust

For the purpose of this thesis, the following definition of trust is adopted: "the will-ingness of one person to increase his or her vulnerability to the actions of another person whose behaviour he or she could not control" (Kim et al., 2005). Trust is an expectation that the other party one chooses to trust, will not behave opportunisti-cally by taking advantage of a given situation, such as a business transaction. It is the belief that the other party will behave in a dependable, ethical, socially appropriate manner and deals with the belief that the trusted party will fulfil its commitments despite the trusting party’s dependence and vulnerability (Gefen, Karahanna, and Straub, 2003).

2.2.2 Online trust

This thesis focusses on online trust. Just like in offline interactions, the party with whom is interacted has the burden of presenting themselves as trustworthy. Accord-ing to Beldad, De Jong, and Steehouder (2010), online organizations must therefore work to improve their reputation, performance, and appearance, where appearance corresponds to e.g. the design of the website. A potential customer can then assess the trustworthiness of a vendor based on the criteria of competence (ability of the vendor to do what is needed), benevolence (vendor to act as expected and accord-ing to interests) and integrity (vendor’s honesty and promise keepaccord-ing) (Beldad, De Jong, and Steehouder, 2010; McKnight, Choudhury, and Kacmarc, 2002). According to McKnight, Choudhury, and Kacmarc (2002), a vendor who possesses these traits is very desirable as an exchange partner, because "he/she will behave ethically, kindly, skilfully, and consistently in the exchange".

The focus of this thesis is on creating a trustworthy mobile website. An organiza-tion’s website could be viewed as a store from the standpoint of building customer trust. Bart et al. (2005) state: "A customer’s interaction with a store is some-what similar to his or her interaction with a website, and consumers develop perceptions of trust in a website based on their interactions with the site". If a consumer gets a positive impression of a website, he or she develops trust with that website. A consumer’s perception of a website’s competence and good intentions contribute to that positive impression. Bart et al. (2005) conclude: "thus, online trust includes consumer perceptions of how the site would deliver on expectations, how believ-able the site’s information is, and how much confidence the site commands". Many antecedents drive these perceptions.

2.3 Reviewing the literature

Now that the concepts of mobile commerce and trust have been explored briefly, the following section provides an overview of which literature was used to find potential determinants of online trust.

(9)

2.3.1 Deciding the scope

It was decided that the literature used to find trust antecedents only stems from the top journals. In this thesis, a journal is considered a top journal if its rank falls within the first quartile (Q1) of its category, based on SCImago Journal Rank indica-tor. On their website, it is stated that the indicator expresses "the average number of weighted citations received in the selected year by the documents published in the selected journal in the three previous years, –i.e. weighted citations received in year X to documents published in the journal in years X-1, X-2 and X-3" (SCImago Journal & Country Rank, 2017). With this scope, only papers are used that the cover key lit-erature sufficiently, have a respectably large field sample and use research questions that are non-trivial. These are examples of standards that journals use to accept or reject a paper for publication (Straub, 2009).

2.3.2 Finding the appropriate literature

Although m-commerce has its own unique features, such as e.g. the interaction style that is unique due to the constraints of the device (screen size) and the usage patterns also differ from those of traditional desktop computers (Feng, Hoegler, and Stucky, 2006), it is essentially still a website. Therefore, it is argued that next to existing m-commerce literature on the topic of trust, e-commerce literature can be used as well.

Literature was searched for via search engines such as Google Scholar and Web of Science. An example query that was used to find appropriate literature is "trust AND mobile commerce OR m-commerce OR electronic commerce OR e-commerce". Then, the journal was reviewed to see whether it would fall within the defined scope. Lastly, the determinants of online trust that had proven to be significant were judged on the applicability to the m-commerce context and scope of the thesis.

TABLE2.1: Literature overview

Study Journal Determinants of online trust

Davis (1989) MIS Quarterly Perceived usefulness & ease of use Koufaris and

Hampton-Sosa (2004) Information & Management

Perceived usefulness & ease of use, customization

Bart et al. (2005) Journal of Marketing Perceived usefulness & ease of use, content

Beldad, De Jong, and

Steehouder (2010) Computers in Human Behavior

Perceived ease of use, customization & personalization, security, content Li and Yeh (2010) Computers in Human Behavior Customization Siau and Shen (2003) Communications of the ACM Customization, design Srinivasan, Anderson,

and Ponnavolu (2002) (2002)

Journal of Retailing Customization

Nilashi et al. (2015) Journal of Retailing and Consumer

Services Security, design, content Pan and Zinkhan (2006) Journal of Retailing Security

Liao, Palvia, and Lin (2006)

International Journal of Information

Management Design, content McKnight, Choudhury,

and Kacmarc (2002)

Journal of Strategic Information

(10)

The studies cited in this thesis, based on the above description of reviewing the literature, can be found in table 2.1.

2.4 Trust antecedents in mobile commerce

Based on the literature as depicted in table 2.1, the following sections give an overview of possible determinants of online trust in the mobile commerce context.

2.4.1 Usefulness and ease of use

Perceived usefulness and ease of use are perceptions taken from the Technology Acceptance Model (TAM) (Davis, 1989). According to Koufaris and Hampton-Sosa (2004), TAM has long been considered a robust model for understanding when users decide to adopt technology, see e.g. Gefen, Karahanna, and Straub (2003), Pavlou (2003), Beldad, De Jong, and Steehouder (2010), and Li and Yeh (2010). Koufaris and Hampton-Sosa (2004) state that since a website is a particular type of technology, it is expected that the two variables have an impact on customer trust beliefs online. In this context, perceived usefulness is defined as "a subjective perception by the customer regarding the site’s utility in his or her shopping task" and perceived ease of use as "the subjective perception by the customer regarding the amount of effort necessary to learn and use the web site" (Koufaris and Hampton-Sosa, 2004).

The perceived ease of use is directly related to the navigation of the layout and the possible sequence of clicks, images and paths on a website (Bart et al., 2005). Factors such as navigation, convenience, usefulness and ease of use drive trust-worthiness (Bart et al., 2005). According to Koufaris and Hampton-Sosa (2004), a well-designed website that is useful and easy to use can be seen as proof of the com-pany’s capabilities. Koufaris and Hampton-Sosa (2004) describe this antecedent by comparing an easy to use website to a store. Customers are more likely to perceive the company as one that is capable of serving them when it has a well organized and laid out store. In contrast, when a customer enters a store that is badly laid out and not well organized, it diminishes trust in that company (Koufaris and Hampton-Sosa, 2004). Believing that the company has the resources and capabilities to fulfil it’s promises, is often described as one of the key antecedents of customer trust in a company (Koufaris and Hampton-Sosa, 2004).

2.4.2 Customization and personalization

Customization and personalization essentially mean the same thing, namely the ex-tent to which an organization’s website can recognize a customer and then tailor its products, services and shopping experience for that customer (Beldad, De Jong, and Steehouder, 2010). This results in an increase in the probability that customers will find something that they may want to buy (Srinivasan, Anderson, and Ponnavolu, 2002). Results of research by Koufaris and Hampton-Sosa (2004), show that the willingness of online organizations to customize their products and services were significant antecedents to trust. They argue that because organizations are able to customize and personalize, customers perceive the organizations as more capable to serve them (Koufaris and Hampton-Sosa, 2004). A study by Li and Yeh (2010) also found that a strong influence of customization on trust in mobile commerce (m-trust). Lastly, Siau and Shen (2003) suggested that personalization of a website can enhance m-trust.

(11)

2.4.3 Security

Nilashi et al. (2015) explain security through four sub criteria, namely security fea-tures, privacy policy statements, payment systems security and site authentication. They found that the security of a website is more important than the other criteria they measured, which are design and content. They also found that a very high level of trust is perceived when the customers feel the security of a website is high. These same findings are mentioned by Beldad, De Jong, and Steehouder (2010), who state that online organizations should seriously consider including strong privacy statements and security features to earn customers’ trust. It is also mentioned that although most customers do not bother to consult or read privacy statements, the mere presence of it would be sufficient to persuade customers to trust the organiza-tion (Pan and Zinkhan, 2006).

2.4.4 Design

According to Nilashi et al. (2015), website design has been defined as "aesthetic fea-tures, manifestation, organization as well as arrangement of online website which are visually pleasing and attractive". According to Liao, Palvia, and Lin (2006), the appearance of a website is an obvious cue by which consumers can assess a retailer’s trustworthiness. If a site is aesthetically pleasing, well-organized and attractive, con-sumers will think that the retailer is willing to invest and regard them as trustwor-thy (Liao, Palvia, and Lin, 2006). According to McKnight, Choudhury, and Kacmarc (2002) it also helps create the impression of general vendor competence (Siau and Shen, 2003).

2.4.5 Content and information quality

According to Nilashi et al. (2015), content of websites is a usability challenge for mobile commerce. Content is what a mobile website presents, namely the offering, appeal, multimedia mix and content type (Nilashi et al., 2015). It is similar to the "information quality" concept discussed in the literature (Liao, Palvia, and Lin, 2006) This content should be free of errors and mistakes in response to consumers’ actions on that site (Bart et al., 2005). Consumers expect a site to not have errors and it should therefore contain correct, current, and complete information and adhere to the rules of correct spelling, grammar, and syntax (Beldad, De Jong, and Steehouder, 2010). According to Liao, Palvia, and Lin (2006), this is because "since customers are not in the position to touch and feel the item in online shopping, they require detailed and clear information to decide on the purchase".

(12)

2.5 Conceptual model

(13)

10

Chapter 3

Methodology

The research objective of this thesis is to find which website factors affect consumer trust in a mobile commerce context in the Netherlands. It was hypothesized that consumers who use a mobile commerce website that utilizes perceived usefulness, ease of use, customization, security features, a visually pleasing design and detailed content have more trust in it than when using a website that does not contain these factors. This study used a quantitative approach to answer the posed research ques-tion. The next sections describes the methodology used.

3.1 Research design

To test the research model, an online questionnaire was used to gather research data. An online questionnaire was chosen because it provides an efficient way of collecting responses from a large sample prior to quantitative analysis (Saunders, Lewis, and Thornhill, 2009).

Respondents were randomly assigned to one of four groups using the block ran-domization in Qualtrics. This feature ensures that every respondent has the same odds to be allocated to a certain condition. Per group, respondents were given three constructs in either a positive or negative version. The constructs were ease of use, se-curity and content for one group and usefulness, customization and design for the other group. It was decided to give a respondent three constructs for two reasons:

1. Saunders, Lewis, and Thornhill (2009) state that the length of a questionnaire has a very high impact on the response rate. Dividing the groups into two would make the questionnaire too long;

2. Creating more than four groups would require even more respondents, since the aim was to get at least 30 respondents per group (as this makes analy-sis easier; according to Saunders, Lewis, and Thornhill (2009), "statisticians have shown that a sample size of 30 or more will usually result in a sampling distribution for the mean that is very close to a normal distribution"). It was expected that getting a minimum of 120 (30 per group) respondents would already be difficult.

Respondents were asked to imagine that they really liked the product depicted and considered ordering. Each construct had its own website depicted in the form of a screenshot (see appendix B). In the positive version, respondents were given a website that utilized that construct, based on the definition given in the literature. In the negative version, respondents were given a website that did not utilize the construct or did so in an obvious, bad way. The screenshots were carefully designed in order to make sure that no other elements could influence trust in that website

(14)

Chapter 3. Methodology 11

(the website brand names/logos were removed, for example). Following questions about that construct, respondents had to answer questions about trusting that web-site.

A pilot was used to validate and evaluate the overall interpretability and clarity of the questionnaire.

3.2 Measures

All measurements of the constructs were adopted from previous literature in order to ensure survey content validity. The only measures that could not be found in existing literature, were for the content construct. For this construct, measures were created based on the definition of the construct found in the literature.

Depending on the group, a total of 25 or 26 questions are included in the ques-tionnaire and are seven or five-point Likert-scale questions, depending on the source of the construct measurement. The following table gives an overview of the con-struct, scale and source.

TABLE3.1: Construct sources

Construct Scale Source

Ease of use 7-point Likert Venkatesh, Ramesh, and Massey (2003) Usefulness 7-point Likert Venkatesh, Ramesh, and Massey (2003) Customization 5-point Likert Koufaris and Hampton-Sosa (2004) Security 7-point Likert Koufaris and Hampton-Sosa (2004) Design 5-point Likert Li and Yeh (2010)

Trust 5-point Likert Li and Yeh (2010) Content 5-point Likert

Since issues can arise from translating measurements (Saunders, Lewis, and Thorn-hill, 2009), measurements were not translated to Dutch even though the target group was Dutch citizens.

3.3 Sample

The questionnaire was distributed through social media and website forums. Since there are no limitations or qualifications of the participants regarding demographics, apart from being a Dutch citizen, the survey was appropriate for all types of respon-dents. Respondents were given an incentive to fill out the survey in the form of a gift card.

(15)

12

Chapter 4

Data analysis and results

4.1 Exploring the sample characteristics

To explore the sample, different descriptive statistics were collected through SPSS. The results can be found in appendix A.1 in the frequency tables and pie charts.

4.1.1 Total sample characteristics

Initially, 146 responses were gathered. Since 23 responses were not fully completed, they were removed from the sample. The total sample (N=123) contains both men (N=87) and women (N=36) and all of the respondents are Dutch citizens.

The sample consists of seven different age groups: 18-24 (N=19), 25-34 (N=48), 35-44 (N=29), 34-54 (N=12), 55-64 (N=13), 65-74 (N=1) and 75-84 (N=1). Another characteristic explored is the education background of the respondents. The sample consists of four different education groups: high school (N=24), bachelors (N=47), masters (N=50) and PhD (N=2).

4.2 Reliability analysis

The scale items should consistently reflect the constructs they are measuring (Field, 2013). To assess the reliability of the six constructs, the reliability analysis within SPSS was utilized. More specifically, this analysis measured the scale reliability through the computation of Cronbach’s α (Field, 2013). Since none of the items are reverse-phrased, no additional steps needed to be taken to deal with a distorted Cronbach’s α.

For this analysis, the positive and negative groups of each construct were put together, resulting in groups of N=61 or N=62. The trust construct was analysed for the complete group (N=123).

Ease of use For this construct, Cronbach’s α was computed at .955, based on the four respective scale items. None of the items led to a higher α if deleted, with the highest potential being Cronbach’s α = .950 (see appendix A.2.1). Thus, the items measuring ease of use can be considered reliable.

Security For this construct, Cronbach’s α was computed at .899, based on the four respective scale items. None of the items led to a higher α if deleted, with the highest potential being Cronbach’s α = .896 (see appendix A.2.2). Thus, the items measuring security can be considered reliable.

(16)

Chapter 4. Data analysis and results 13

Content For this construct, Cronbach’s α was computed at .948, based on the three respective scale items. None of the items led to a higher α if deleted, with the highest potential α being .927 (see appendix A.2.3). Thus, the items measuring content can be considered reliable.

Usefulness For this construct, Cronbach’s α was computed at .940, based on the four respective scale items. None of the items led to a higher α if deleted, with the highest potential α being .938 (see appendix A.2.4). Thus, the items measuring usefulness can be considered reliable.

Customization For this construct, Cronbach’s α was computed at .862, based on the three respective scale items. None of the items led to a higher α if deleted, with the highest potential α being .863 (see appendix A.2.5). Thus, the items measuring customization can be considered reliable.

Design For this construct, Cronbach’s α was computed at .946, based on the three respective scale items. None of the items led to a higher α if deleted, with the highest potential α being .943 (see appendix A.2.6). Thus, the items measuring design can be considered reliable.

Trust For this construct, Cronbach’s α was computed at .945, based on the four re-spective scale items. None of the items led to a higher α if deleted, with the highest potential α being .943 (see appendix A.2.7). Thus, the items measuring trust can be considered reliable.

According to Field (2013), a Cronbach’s α ranging from .7 to .8 is perceived as an acceptable value and .9 as excellent. Thus, all scale items from each of the seven constructs appear to be reliable measurements, because the α is higher than .7 for every construct.

4.3 Group sample characteristics

The sample was randomly divided into four groups. Each group is considered a separate experiment and is therefore analysed individually. According to Saunders, Lewis, and Thornhill (2009), "statisticians have shown that a sample size of 30 or more will usually result in a sampling distribution for the mean that is very close to a normal distribution". Since two of the groups are below that threshold, the group samples will be checked on whether they are significantly different from a normal distribution or not.

The SPSS output of the group characteristics can be found in appendix A.3.

Group 1 The first group, with the constructs ease of use, security and content in a neg-ative form, had a size of N=33. To test whether the sample is normally distributed, a Shapiro-Wilk analysis was utilized. According to Field (2013), if the test is significant (p <.05) then the distribution is significantly different from a normal distribution (i.e. it is non-normal). A Shapiro-Wilk analysis reveals that none of the constructs are significantly different from a normal distribution.

(17)

Group 2 The second group, with the constructs ease of use, security and content in a positive form, had a size of N=29. A Shapiro-Wilk analysis reveals that the con-structs ease of use, D(29) = .047, p < .05, and content, D(29) = .000, p <.001, were both significantly different from a normal distribution. Also both the ease of use and con-tent trust items were significantly different, with D(29) = .016, p <.05 and D(29) = .009, p <.05 respectively.

Group 3 The third group, with the constructs usefulness, customization and design in a negative form, had a size of N=27. A Shapiro-Wilk analysis reveals that none of the constructs are significantly different from a normal distribution.

Group 4 The last group, with the constructs usefulness, customization and design in a positive form, had a size of N=34. A Shapiro-Wilk analysis reveals that all of the constructs are significantly different from a normal distribution.

4.4 Model validity

Due to the non-normal distribution of most of the constructs, bivariate correlation analysis was used to measure whether the six constructs are actually able to predict the outcome variable (trust) in a significantly good degree. This analysis is required before the results of the groups can be compared, because if there is no relationship between the constructs and trust, it does not make sense to compare the trust scores of the four groups.

Although Spearman’s statistics is the most popular of the coefficients, Field (2013) states that "there is much to suggest that Kendall’s statistic is actually a better esti-mate of the correlation in the population" (Field, 2013). Field (2013) also states that Kendall’s tau is better for small samples. As such, more accurate generalizations can be drawn. Therefore it was decided that Kendall’s tau (τ ) was going to be used. An overview of the SPSS output can be found in appendix A.4.

A Kendall’s τ rank-order correlation was run to determine the relationship be-tween trust and each construct. An overview of the results can be found in table 4.1.

TABLE4.1: Model validity results

Construct Result Ease of use rτ (29) = .544, p = .000 Security rτ (29) = .630, p = .000 Content rτ (29) = .713, p = .000 Usefulness rτ (34) = .729, p = .000 Customization rτ (34) = .728, p = .000 Design rτ (34) = .574, p = .000

The results of the analysis show that for each construct, a significant (p < .001) correlation coefficient was found. This means that a correlation exists between each of the constructs and trust. All of the correlation coefficients are positive, meaning that if e.g. ease of use increased, trust increased also.

(18)

4.5 Results

Due to the non-normal distribution of the groups, a Mann-Whitney-U test was used instead of a T-test for hypothesis testing. Trust scores between the positive and neg-ative groups were analysed (e.g. ease of use trust positive and ease of use trust negative). The following section gives an overview of the results of this test. The SPSS output can be found in appendix A.5.

Hypothesis 1 Trust in the positive group (Mdn = 24.00) did differ significantly from the negative group (Mdn = 14.00) with regard to the ease of use construct, U = 31.00, z = -6.329, p <.001, r = -.80. This means that the first (null) hypothesis, there is no significant difference in trust in a mobile commerce website that is easy to use compared to one that is not, is rejected.

The effect size (r = -.80) shows a huge effect for the positive group, since the effect size is well above the .5 threshold for a large effect (Field, 2013).

Hypothesis 2 Trust in the positive group (Mdn = 15.00) did not differ significantly from the negative group (Mdn = 15.00) with regard to the security construct, U = 388,50, z = -1,275, ns, r = -.16. This means that the second (null) hypothesis, there is no significant difference in trust in a mobile commerce website that is secure compared to one that is not, is not rejected.

Hypothesis 3 Trust in the positive group (Mdn = 24.00) did differ significantly from the negative group (Mdn = 14.00) with regard to the content construct, U = 131.00, z = -4,916, p <.001, r = -.62. This means that the third (null) hypothesis, there is no significant difference in trust in a mobile commerce website that has good content compared to one that does not, is rejected.

The effect of content on trust between the two groups also represent a large effect, since the effect size (r = -.62) is above the .5 threshold for a large effect (Field, 2013).

Hypothesis 4 Trust in the positive group (Mdn = 22.00) did differ significantly from the negative group (Mdn = 14.00) with regard to the usefulness construct, U = 148.50, z = -4,521, p <.001, r = -.58. This means that the fourth (null) hypothesis, there is no significant difference in trust in a mobile commerce website that is perceived useful compared to one that is not, is rejected.

The effect of the perceived usefulness on trust between the two groups also rep-resent a large effect, since the effect size (r = -.58) is above the .5 threshold for a large effect (Field, 2013).

Hypothesis 5 Trust in the positive group (Mdn = 22.50) did differ significantly from the negative group (Mdn = 13.00) with regard to the customization construct, U = 155.00, z = -4,429, p <.001, r = -.57. This means that the fifth (null) hypothesis, there is no significant difference in trust in a mobile commerce website that utilizes customization compared to one that does not, is rejected.

The effect of customization on trust between the two groups also represent a large effect, since the effect size (r = -.57) is above the .5 threshold for a large effect.

Hypothesis 6 Trust in the positive group (Mdn = 22.00) did differ significantly from the negative group (Mdn = 15.00) with regard to the design construct, U = 176.00, z = -4,123, p <.001, r = -.53. This means that the sixth (null) hypothesis, there is no

(19)

significant difference in trust in a mobile commerce website that has a good design compared to one that does not, is rejected.

The effect of the design on trust between the two groups also represent a large effect, since the effect size (r = -.53) is above the .5 threshold for a large effect (Field, 2013).

(20)

17

Chapter 5

Conclusion and limitations

5.1 Conclusion

In this study, customer trust in a mobile commerce website was investigated from the literature. More specifically, which website factors influence customer trust dur-ing online shoppdur-ing on a smartphone. Literature on the subject has shown that there are several trust antecedents in mobile commerce websites, namely usefulness, ease of use, customization, security, design and content. While these constructs have shown to be determinants of trust in several countries, they were not yet researched in the Netherlands. Therefore, the research question of this thesis was "What are the factors that affect customer trust in a mobile commerce website in the Netherlands?". Based on existing literature, a model was developed that incorporated these trust antecedents regarding a website. The results show that usefulness, ease of use, cus-tomization, design and content were all significant antecedents to trust in a mobile commerce website. Finally, no support was found for the hypothesis that security is an antecedent to trust in a mobile commerce website.

This study contributes to the literature, since the six constructs were not yet re-searched in the Netherlands. Findings from this study are partly in line with those of Li and Yeh (2010) and Nilashi et al. (2015), which are both from different countries than the subject of this study. The findings of this study, however, seem to suggest that Dutch consumers do not trust a mobile commerce website with security fea-tures more than a website that does not contain these feafea-tures. This is a different finding than Nilashi et al. (2015), because in their study the security component was the most important construct (even more so than design and content). As stated in the introduction, trust antecedents may differ from country to country (Vance, Elie-Dit-Cosaque, and Straub, 2008). This difference in preference with regard to trust antecedents could be an explanation of the discrepancy between research results.

The practical implications of this study are that online merchants should focus their efforts on improving their mobile commerce website by making use of the five trust antecedents that were found significant. Even though Dutch consumers have a slight preference for avoiding uncertainty (Hofstede, 2010), consumers don’t seem to avoid uncertainty by looking at security features of a website. Therefore, no effort should be put in making a potential customer aware of the security features in order to gain their trust.

5.2 Limitations

Although this research has reached its aims, there were some limitations that need to be considered.

(21)

Chapter 5. Conclusion and limitations 18

Due to the way the questionnaire was set-up, each group was treated as an in-dividual experiment. It was therefore not possible to analyse the results as one ex-periment and e.g. make use of multiple regression to see which of the constructs impacted trust the most. In a future study, the questionnaire could be set-up differ-ently in order to make this possible.

Because of the non-normal distribution of the groups, non-parametric alterna-tives had to be used for the analysis. Field (2013) states that by using non-parametric tests, some information about the magnitude of differences between the scores is lost. As a result, non-parametric tests can be less powerful than their parametric counterparts (Field, 2013). If the sample was normally distributed, parametric tests could have been used instead. This could have been overcome by a larger sample size, because of the central limit theorem (Field, 2013).

Due to convenience sampling, the sample may have been prone to bias or over/under-representation of particular groups (Saunders, Lewis, and Thornhill, 2009). As the sample characteristics show, most of the respondents had a masters’ degree (N=50). This is not a right representation of the total Dutch population, because in reality, only roughly 10% of the population has a masters’ degree (CBS, 2017).

Results may also be influenced by the research environment. Pengnate and Sarathy (2017) conducted a study by using an experimental apartment rental web-site. They state that since the website was experimental, observed behaviour may have been different from behaviour on real websites since there was no actual risk of using the website. This is also the case in this study, since the websites were depicted in the form of a screenshot.

(22)

(23)

Appendix A. Appendix 20

Appendix A

Appendix

(24)

A.2 Reliability SPSS output

(25)

(26)

(27)

(28)

(29)

(30)

(31)

A.3 Group sample characteristics SPSS output

A.3.1 Group 1

(32)

A.3.3 Group 3

(33)

A.4 Model validity SPSS output

A.4.1 Ease of Use

A.4.2 Security

(34)

A.4.4 Usefulness

A.4.5 Customization

(35)

A.5 Results SPSS output

A.5.1 Ease of Use

(36)

A.5.3 Content

(37)

A.5.5 Customization

(38)

35

Bibliography

Bart, Yakov et al. (2005). “Are the Drivers and Role of Online Trust the Same for All Web Sites and Consumers? A Large-Scale Exploratory Empirical Study”. In: Journal of Marketing 69.4, pp. 133–152.

Beldad, Ardion, Menno De Jong, and Michaël Steehouder (2010). “How shall i trust the faceless and the intangible? A literature review on the antecedents of online trust”. In: Computers in Human Behavior 26.5, pp. 857–869.

CBS (2017). CBS StatLine - Bevolking; hoogst behaald onderwijsniveau; geslacht, leeftijd en herkomst.

Chong, Alain Yee Loong, Felix T S Chan, and Keng Boon Ooi (2012). “Predicting consumer decisions to adopt mobile commerce: Cross country empirical exam-ination between China and Malaysia”. In: Decision Support Systems 53.1, pp. 34– 43.

Cyr, Dianne (2013). “Website design, trust and culture: An eight country investiga-tion”. In: Electronic Commerce Research and Applications 12.6, pp. 373–385.

Cyr, Dianne, Milena Head, and Alex Ivanov (2006). “Design aesthetics leading to m-loyalty in mobile commerce”. In: Information and Management 43.8, pp. 950– 963.

Davis, F. D. (1989). “Perceived usefulness, perceived ease of use, and user acceptance of information technology”. In: MIS Quarterly, pp. 319–340.

Feng, Haiqi, Tamara Hoegler, and Wolffried Stucky (2006). “Exploring the Critical Success Factors for Mobile Commerce”. In: International Conference on Mobile Busi-ness (ICMB’06) November 2015, p. 40.

Field, Andy (2013). Discovering Statistics Using SPSS.

Gefen, David, Elena Karahanna, and Detmar W Straub (2003). “Trust and TAM in Online Shopping: An Integrated Mode”. In: MIS Quarterly 27.1, pp. 51–90. Hofstede, Geert (2010). Cultural Dimensions - Taiwan in comparison with Malaysia. Junglas, Iris and Richard T Watson (2006). “The U-Constructs : Four Information

Drives”. In: Communications of the Association for Information Systems 17.1, pp. 569– 592.

Kim, Dan J. et al. (2005). “A multidimensional trust formation model in B-to-C e-commerce: A conceptual framework and content analyses of academia/practitioner perspectives”. In: Decision Support Systems 40.2, pp. 143–165.

Koufaris, Marios and William Hampton-Sosa (2004). “The development of initial trust in an online company by new customers”. In: Information and Management 41.3, pp. 377–397.

Li, Yung Ming and Yung Shao Yeh (2010). “Increasing trust in mobile commerce through design aesthetics”. In: Computers in Human Behavior 26.4, pp. 673–684. Liao, Chechen, Prashant Palvia, and H.N. Lin (2006). “Liao, C., Palvia, P., and Lin,

H. N. The roles of habit and web site quality in e-commerce.” In: International Journal of Information Management, 26.6, pp. 469–483.

(39)

BIBLIOGRAPHY 36

McKnight, D. H., Vivek Choudhury, and C Kacmarc (2002). “The impact of ini-tial consumer trust on intentions to transact with a web site: A strust building model.” In: Journal of Strategic Information Systems 11, pp. 297–323.

Ngai, E. W T and A. Gunasekaran (2007). “A review for mobile commerce research and applications”. In: Decision Support Systems 43.1, pp. 3–15.

Nilashi, Mehrbakhsh et al. (2015). “The role of Security, Design and Content factors on customer trust in mobile commerce”. In: Journal of Retailing and Consumer Ser-vices 26, pp. 57–69.

Ofcom (2016). Communications Market Report 2016. Tech. rep. Ofcom.

Pan, Yue and George M. Zinkhan (2006). “Exploring the impact of online privacy disclosures on consumer trust”. In: Journal of Retailing 82.4, pp. 331–338.

Pavlou, Paul a. (2003). “Consumer Acceptance of Electronic Commerce: Integrating Trust and Risk with the Technology Acceptance Model”. In: International Journal of Electronic Commerce 7.3, pp. 69–103.

Pengnate, Supavich (Fone) and Rathindra Sarathy (2017). “An experimental investi-gation of the influence of website emotional design features on trust in unfamiliar online vendors”. In: Computers in Human Behavior 67, pp. 49–60.

Saunders, Mark, Philip Lewis, and Adrian Thornhill (2009). Research methods for busi-ness students. Fifth edition. Pearson Education, pp. 0–264.

SCImago Journal & Country Rank (2017). Scientific Journal Rankings.

Siau, Keng and Zixing Shen (2003). “Building Customer Trust in Mobile Commerce”. In: Communications of the ACM 46.4, pp. 91–94.

Srinivasan, Srini S., Rolph Anderson, and Kishore Ponnavolu (2002). “Customer loy-alty in e-commerce: An exploration of its antecedents and consequences”. In: Journal of Retailing 78.1, pp. 41–50.

Straub, Detmar W. (2009). “Why Top Journals Accept Your Paper”. In: MIS Quarterly 33.3, pp. 262–265.

Thuij, Gino (2016). Thuiswinkel Markt Monitor Q3 2016.

Vance, Anthony, Christophe Elie-Dit-Cosaque, and Detmar W. Straub (2008). “Exam-ining Trust in Information Technology Artifacts: The Effects of System Quality and Culture”. In: Journal of Management Information Systems 24.4, pp. 73–100. Venkatesh, Viswanath, V. Ramesh, and Anne P. Massey (2003). “Understanding

us-ability in mobile commerce”. In: Communications of the ACM 46.12, p. 53.

Wong, Yuk Kuen and Chao Jung Hsu (2008). “A confidence-based framework for business to consumer (B2C) mobile commerce adoption”. In: Personal and Ubiq-uitous Computing 12.1, pp. 77–84.

Wu, Jen Her and Shu Ching Wang (2005). “What drives mobile commerce? An em-pirical evaluation of the revised technology acceptance model”. In: Information and Management 42.5, pp. 719–729.

(40)

(41)

(42)

(43)

(44)

(45)

(46)

(47)

(48)

(49)

(50)

(51)

(52)

(53)

(54)

(55)

(56)

(57)

(58)

(59)

(60)

(61)

(62)

(63)

(64)