Unified Theory of Acceptance and Use of Technology applied to mobile Augmented Reality Applications in The Netherlands for retail purposes
Name: Ghislaine Heidman
Student number: 10813543
Thesis supervisor: Dhr. Dr. Y.B. Altayligil Study: MSc Business Administration
Track: Digital Business
Institute: University of Amsterdam Date of submission: 22.06.2018
Statement of originality
This document is written by Ghislaine Heidman who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3 Abstract
New digital technologies continuously emerge that could give retailers opportunities to increase sales. One of these emerging digital technologies is augmented reality (AR), which has been available on mobile devices since 2017. Thus, it is not surprising that retailers are acting on these new possibilities and have started developing their own mobile AR-based applications. However, there is limited research on the acceptance of mobile AR apps in retail. Therefore, this study investigates a revised Unified Theory of Acceptance and Use of Technology (UTAUT2) model in the context of mobile AR retail applications and provides an understanding of the different drivers that influence the behavioural intention to use mobile AR apps in the retail industry, specifically for the Dutch customer. Furthermore, the
moderating effect of personal innovativeness in the domain of information technology is investigated. The data was collected via an online survey (n = 179). The proposed model was analysed via a hierarchical linear regression model, a stepwise regression model and process analysis. From the 13 hypotheses, 4 were found to be statistically significant. The results indicate that personal innovativeness in the domain of information technology does not moderate the relationship between the drivers and the behavioural intention to use AR apps in retail. However, personal innovativeness in the domain of information technology is a driver of the behavioural intention to use mobile AR apps in retail. More drivers that influence the behavioural intention to use AR apps in retail are performance expectancy, social influence and hedonic motivation. The findings of this study provide new insights for both researchers and retailers. For retailers, these insights indicate they should aim to develop the AR-based applications according to consumer’s hedonic motivation, convince the customer of the apps’ benefits and focus on early adaptors to drives m-commerce sales.
Key words: Augmented Reality, Mobile Applications, UTAUT2, Technology Acceptance, Consumer Acceptance, Retail
Table of Contents
Abstract 3
1. Introduction 8
1.1 Problem Statement 9
1.2 Research Contributions & Research Questions 10
1.3 Thesis Outline 11
2. Literature Review 12
2.1 Technology 12
2.1.1 Augmented Reality 12
2.1.2 Virtual Reality 13
2.1.3 Retail Segments Using Mobile AR Applications 13
2.2 Technology Acceptance Models 15
2.2.1 Technology Acceptance Model 15
2.2.2 UTAUT: Unified Theory of Acceptance and Use of Technology 16 2.2.3 UTAUT2: Unified Theory of Acceptance and Use of Technology 2 17
2.3 Personal Innovativeness in the domain of IT 18
2.3.1 Personal Innovativeness in the domain of IT as Moderator 19
2.3.2 Personal Innovativeness in the domain of IT as a predictor of UTAUT2 20
2.4 Conceptual Framework 20
2.5 Hypotheses 22
3. Data & Method 26
3.1 Methodology 26
3.2 Data collection and sample 27
4. Results 30
4.1 Preliminary steps 30
5
4.2.1 Principal Components Analysis 31
4.2.2 Convergent validity 34
4.2.3 Correlation Matrix 34
4.3 Analyses 37
4.3.1 Hierarchical Regression Model 37
4.3.2 Stepwise Linear Regression Model 39
4.3.2 Moderation effect 42
4.3.3 One-way ANOVA 44
4.4 Hypotheses testing 47
5. Discussion 49
5.1 Findings Confirm UTAUT 2 49
5.2 Role of Personal Innovativeness in the Domain of IT 50
5.3 Retail segments 51
5.4 Contribution to the theory 51
5.5 Managerial implications 52
5.6 Limitations and future research 53
6. Conclusion 54
References 55
List of Tables
Table 1. Retail segments using Augmented Reality for M-commerce 14
Table 2. UTAUT2 survey items (Venkatesh et al., 2012) 27
Table 3. Educational Level 29
Table 4. Age Groups 29
Table 5. Reliability, Cronbach’s Alpha 31
Table 6. KMO and Bartlett’s Test 32
Table 7. Factor Loading PCA 33
Table 8. Mean, Standard Deviation and Correlations 36
Table 9. Hierarchical Regression Model of Behavioral Intention 39
Table 10. Stepwise regression – Model Summary 40
Table 11. Stepwise regression – Coefficients 41
Table 12. Process moderation effect of PIIT on performance expectancy 43 Table 13. Process moderation effect of PIIT on effort expectancy 43 Table 14. Process moderation effect of PIIT on social influence 43 Table 15. Process moderation effect of PIIT on hedonic motivation 44
Table 16. Process moderation effect of PIIT on habit 44
Table 17. One-way ANOVA Retail segments 45
Table 18. One-way ANOVA Gender 46
Table 19. One-way ANOVA Education 46
7 List of Figures
Figure 1. Technology Acceptance Model 15
Figure 2. Unified Theory of Acceptance and Use of Technology 17 Figure 3. Unified Theory of Acceptance and Use of Technology 2 18
Figure 4. Personal Innovativeness in the domain of IT 19
Figure 5. Conceptual Model Behavioral Intention to Use Augmented Reality 21
Figure 6. Histogram model 5 42
Figure 7. Path Coefficients and significance level 48
Abbreviations AR: Augmented Reality
VR: Virtual Reality Apps: Applications
PIIT: Personal Innovativeness in de domain of IT PE: Performance Expectancy
EE: Effort Expectancy SI: Social Influence HM: Hedonic Motivation FC: Facilitating Conditions
H: Habit
BI: Behavioral Intention to Use
UTAUT(2): Unified Theory of Acceptance and Use of Technology (2) TAM: Technology Acceptance Model
1. Introduction
Pokémon Go, the app that went viral in 2016, had more daily active users than Tinder and its average daily usage time was higher than for Whatsapp (Schwartz, 2016). The app allows mobile users to interact with their surroundings via their smartphone camera. Favourite Pokémon can be caught with Pokéballs, and people can use their physical locations to obtain credits in the virtual map. The app combines a live view of the physical world with virtual reality elements, called augmented reality (AR) (Azuma, Behringer, Julier & Macintyre, 2001). Pokémon brought AR to the masses, although the technology had existed for decades. The success of the Pokémon Go app can indicate to businesses consumers’ eagerness to embrace AR in the near future (Chen, 2017). In the world of mobile commerce (m-commerce) especially, the question marketers should ask themselves is ‘Will augmented reality shape the future in the m-commerce environment?’
In m-commerce, one of the main obstacles consumers face is determining whether a product or service is right for them. Often, consumers do not know whether a product or service is in the same style, the same colour and size in reality compared to what they have seen online. Augmented reality gives consumers the opportunity to test and see the product in different settings, such as at home or even on the consumers’ body and face via AR mirror software. Furthermore, AR can be used to give a consumer additional information (online on a smartphone) when looking at the physical product via the phone’s mobile screen (Chen, 2017). For example, the Yummly app, a search engine for food recipes, is now equipped with an AR function that allows users to point a phone’s camera over the items users have in their kitchens. Yummly will recommend recipes that include the ingredients a user already has at home (Yummly, 2018). To conclude, AR could offer consumers confidence that eventually could motivate a purchase decision for a certain product or service (Chen, 2017).
9 Retailers are increasingly implementing AR within their mobile applications to market their products or services (Chen, 2017). For instance, IKEA gives consumers the opportunity to envision IKEA furniture in a physical environment, such as their living room (IKEA, 2017). The application allows consumers to see the different colours and styles of many varieties of furniture, and consumers could, therefore, make a more confident purchase decision. With this application, IKEA distinguishes itself from other furniture retailers. Not only products can be envisioned with AR, but there are also applications that allow consumers to envision a service that could be delivered. Examples of these are tattoo apps, measure kit apps and the app called Modiface. The Modiface application comes with a new sort of AR: the AR mirror software (Modiface, 2018). Mostly used by beauty brands, this software enables consumers to look at a mirror in a store or at the in-front camera of their smartphones and see a product on the consumers’ body, face or hair (Modiface, 2018). Hairdressers could use this type of apps to allow consumers to first see the end result before they start cutting or dyeing hair.
The AR mirror software is also used by retailers such as Ray-Ban to gives consumers the opportunity to see different styles of sunglasses on their faces (Ray Ban, 2018). In
addition, the Tattoo app called Inkhunter allows consumers to see different tattoo designs and sizes on their bodies before they are inked (Inkhunter, 2018).
1.1 Problem Statement
The continuous rise of m-commerce is inevitable and unstoppable according to data (Ismail, 2018). Therefore, AR is at the start of being introduced to m-commerce, and it would be beneficial for retailers to make AR a significant component of their marketing and sales. Ultimately, this integration will give the retailers the opportunity to stay ahead of their competitors.
However, before incorporating AR within a business, retailers should be aware of the factors influencing consumers’ acceptance of this new technology used in marketing.
Therefore, this study aims to measure consumers’ acceptance for augmented reality retail applications by extending the UTAUT2 model with a new moderator, personal innovativeness in the domain of IT. Furthermore this study aims to determine whether these factors are different for different retail segments.
1.2 Research Contributions & Research Questions
Prior research is limited to the acceptance of AR technology in general and has not
empirically tested AR acceptance in retail applications. Therefore, this study will contribute by specifically measuring the factors that influence the Dutch consumers’ acceptance of AR used in retail applications, that will be extended by a new factor/moderator, personal
innovativeness in the domain of IT. In addition, no previous research measures to what extent there are differences between retail segments. Therefore, this research examines four large retail segments and determines whether these drivers differ per segment: Beauty and
Cosmetics, Food, Home Décor and Fashion items. These four segments were chosen based on existing AR mobile apps that aim to sell products or services (Kolo, 2018; Sheehan, 2018). The factors that will influence consumers’ acceptance should help marketers to design augmented reality applications more efficiently in the world of m-commerce.
Furthermore, most prior studies are limited to one particular media environment, computers, because mobile phones have only recently become compatible with the AR function. This is a limitation in literature because AR-based applications are used on mobile phones. For example, in September 2017, Apple launched its ARKit, which is only
compatible with the iOS11 or higher and the Samsung S9 with Bixby Vision AR function. Therefore, this research focuses specifically on customer acceptance on mobile devices, as
11 gathering data about customer acceptance is now possible due to Snapchat filters, the high usage of the Pokémon Go app and other AR related apps.
Despite available data, general literature about factors influencing the Dutch
consumers’ acceptance of AR retail applications seems to be missing; therefore, the following research question was formulated:
‘What are the drivers that influence consumers’ behavioural intention to use mobile augmented reality applications in retail and how does personal innovativeness influence these relationships?’
1.3 Thesis Outline
The remainder of this paper is organized as follows. The second chapter elaborates on the literature found on the following topics: augmented reality, virtual reality and retail segments using mobile AR applications. Furthermore, the technology acceptance theories technology acceptance model, the Unified Theory of Acceptance and Use of Technology (UTAUT) and UTAUT2models from Venkatesh, and the Personal Innovativeness construct from Agarwal and Prasad are discussed (1998). In Chapter 3, the research model and the hypotheses are presented. Furthermore, the modified UTAUT2 model is explained in detail. The fourth chapter discuss the data and method that is used. The fifth chapter, Results, elaborates on the gathered data and consist of the preliminary steps, data analyses, hypotheses testing and an empirical section. In Chapter 6, the discussion is presented, followed by a more explicit conclusion in Chapter 7, which finishes with the limitations and future research of this study.
2. Literature Review
In the first section of this chapter, the literature on the main subject is discussed to develop a background for this study. The first section is divided into three sub-sections: Augmented Reality, Virtual Reality and Retail Segments Using Mobile AR Applications. In the second section, three acceptance models that explain consumer’s acceptance of new technologies are discussed. In the third section, the personal innovativeness model in the domain of IT is presented. Followed by the fourth section where the theoretical framework is presented by showing the conceptual model of this study. The fifth section includes the hypotheses that are tested.
2.1 Technology
2.1.1 Augmented Reality
New digital technologies constantly emerge; one of these is AR. Augmented reality is likely to have an impact on marketing in the near future (Kannan, 2017), although the impact has not yet been scientifically proved. Augmented reality is a new method of visual informatics and is defined by Azuma et al. (2001) as ‘a system that supplements the real world with virtual (computer generated) objects that appear to coexist in the same space as the real world’. In the age of e- and m-commerce, showcasing products are becoming increasingly more important. Augmented Reality creates new possibilities for content delivery to consumers, as the technology allows for online customers to fully inspect a product before they decide to purchase it. Furthermore, the fast developments in the mobile industry have brought AR experiences to mobile devices. Prior studies have shown that augmented reality-based marketing applications are a more persuasive tool for consumers than the traditional web-based product presentation (Yim, Chu & Sauer, 2017).
13 2.1.2 Virtual Reality
It is essential to understand the differences between augmented reality and virtual reality (VR) to better understand the strength of AR and realize how it could be beneficial for businesses. ‘Virtual Reality is a three-dimensional computer-generated environment designed for and controlled by a person’s physical presence; it is a disruption between the physical and virtual world’ (Dictionary, 2018). Unlike VR, AR technology visualizes virtual objects in a real-world environment (Azuma et al., 2001).
2.1.3 Retail Segments Using Mobile AR Applications
Today, a variety of retailers use AR technology to empower their customers, improve customers’ shopping experience and help their products stand out (Sheehan, 2018). The technology is either used for in-store experience or for AR experience at home. Examples of AR in-store experiences are the AR enabled mirror and the AR fitting room (Sheehan, 2018). However, AR mobile applications are more often used for online shopping experiences, although some mobile applications encourage engagement with in-store signage and displays (American Apparel, 2018). Most of the literature is about the acceptance of AR in
entertaining and education context (Balog & Pribeanu, 2010; Sumadio & Rambli, 2010). However, according to Rese et al. AR apps in other contexts, such as fashion and toys, should also be explored with regard to consumers’ acceptance models (2017) as well. As this
research focuses on mobile AR applications, only the retailers who developed a mobile AR-based application will be included (before January 2018). Since mobile phones have been compatible with the AR software for less than a year, not much is written about the different segments. For this reason, four segments were created, and all existing AR apps are listed in Table 1 below (Kolo, 2018; Sheehan, 2018). Only the AR apps that aim to sell products, and therefore belong to retail, were included. Furthermore, a distinction was made between the front camera used to explore AR or the normal phone camera. These are called AR mirror
function and AR view, respectively. Table 1 displays the four retail segments with a
description and the corresponding applications, those were available before January 2018. For every retail segment two applications were used for this study, which are the following: (1) Beauty and Cosmetics; Sephora Virtual Artist and L’Oreal, (2) Food; KabaQ and Yummly, (3) Home Décor; IKEA place and Amazon AR view, (4) Fashion; Topology and Gap. The function AR mirror is indicated with a (M) behind the app name.
Table 1.Retail segments using augmented reality for m-commerce
Retail segment Description Mobile Applications
Beauty and Cosmetics AR mirror software used to try on new make-up styles, and see how different products look on the customer’s face.
Modiface (M), L’Oreal (M), Sephora Virtual Artist (M), Rimmel (M).
Food Yummly (a recipes app) uses the phone’s
camera to catalogue food items a customer has in his or her kitchen and recommends recipes that include the ingredients the customer has at home. The additional products needed for the meal can be ordered via the app.
KabaQ gives restaurant customers the opportunity to see meals in AR before they choosing their dinners.
Yummly, KabaQ
Home Décor The apps allow customers to visualize what new furniture or paint colour might look like in their living rooms.
IKEA place, Dulux Visualizer, Amazon AR view
Fashion These apps allow a customer to visualize a piece of clothes or shoes on their bodies. Converse and Ray-Ban are two of the first apps to use this technology, so the app quality is low. The Gap is still developing its app, and Topology gives people the opportunity to measure the best fit of glasses and see the end result.
Converse, Ray-Ban (M), Gap, Topology (M).
15 2.2 Technology Acceptance Models
2.2.1 Technology Acceptance Model
Understanding consumers’ acceptance of technology has become a central topic in previous research because of the fast developments in information technology. Over the years, many researchers have written theories and proposed models to better understand the acceptance of new technologies. One of the first models to measure consumer acceptance is the technology acceptance model (TAM) from Davis (1989) (shown in Figure 1). The TAM is an information system theory that models how consumers come to accept a new technology and make use of the following variables: perceived usefulness, perceived ease of use, attitude toward using and behavioural intention to use (Davis, 1989).
Existing empirical studies have focused on the TAM (Davis, 1989) when they were evaluating online customers’ acceptance of augmented reality (Rese, Baier, Geyer-Schulz & Schreiber, 2017). The TAM is often modified with external variables, such as perceived informativeness and perceived enjoyment. Both external variables have been tested and are confirmed to have a positive impact on perceived usefulness. However, it was also indicated that the acceptance of AR applications could vary among cultures (Choi, Lee, Sajjad & Lee, 2014; Lee, Chung & Jung 2015). Since the Dutch consumer has not yet been scrutinized and the acceptance of new technologies could vary among cultures, it would be beneficial for Dutch retailers to know the drivers of technology acceptance for Dutch consumers. Figure 1. Technology Acceptance Model (Davis, 1989).
Furthermore, there may be differences between hedonic and functional utility (Rese et al., 2017).
2.2.2 UTAUT: Unified Theory of Acceptance and Use of Technology
Multiple papers describe different models to test the technology acceptance with different determinants. Venkatesh, Moris, Davis & Davis reviewed eight of these user acceptance models (2003). In their research, they formulated and validated a unified model that
integrated all determinants of these eight models (Venkatesh et al., 2003) (shown in Figure 2). In fact, the UTAUT model is a revised version of the TAM. The original UTAUT has four main determinants: performance expectancy (PE), effort expectancy (EE), social influence (SI) and facilitating conditions (FC). These determinants explain 70% of the variance in the behavioural intention to use (Venkatesh, Thong & Xu, 2012).
Performance expectancy is defined as to what extent individuals believe the technology will help them to better perform tasks. This determinant corresponds to the determinant perceived usefulness in the TAM (Venkatesh et al., 2003). Venkatesh et al. indicates also that the PE construct is the strongest predictor of behavioural intention to use (2003). Effort Expectancy corresponds to perceived ease of use in the TAM and relates to what extent the individual believes the technology is easy to use. Social Influence relates to what extent others’ opinions influence an individual’s intention to use the technology, mostly in that individual’s own network of family and friends (Venkatesh et al., 2003). Facilitating Conditions corresponds to what extent an individual believes he or she has the resources to make use of the technology (Venkatesh et al., 2003).
The model also has four different moderators: PE by age and gender; EE by age, gender and experience; SI by age, gender, experience and voluntariness of use and FC by age and experience. In addition, the original study has proved a positive relationship between FC and use behaviour.
17 The original UTAUT model was constructed when investigations into technology acceptance were more concerned with employees in large corporations than with the individual consumer. Therefore, Venkatesh, together with Thong and Xu, developed and empirically tested an extended version of the UTAUT model (2012). This model is called the UTAUT2 and is designed to understand technology acceptance from a consumer’s
perspective.
2.2.3 UTAUT2: Unified Theory of Acceptance and Use of Technology 2
The UTAUT2 model includes three additional determinants, hedonic motivation (HM), price value (PV) and habit (H), which predict consumers’ behavioural intention to use a new technology (Venkatesh et al., 2012) (shown in Figure 3).
According to Venkatesh et al. HM is the extent to which an individual can enjoy a technology and the pleasure derived from using it (2012). This determinant was added to the UTAUT model, as it is an important factor in consumer technology use (Brown & Venkatesh, 2005). The second determinant, PV, relates to the individual’s belief the technology is worth its monetary costs (Brown & Venkatesh, 2005). This determinant was added because the Figure 2, Unified theory of acceptance and use of technology (Venkatesh, Moris, Davis & Davis, 2003).
UTAUT2 model is designed from a consumer’s perspective, and therefore, consumers, rather than an organization, bear the monetary costs of a technology (Venkatesh et al., 2012). The third addition is the determinant H, which relates to the individual's previous use of the technology and the customer’s belief that their use of that particular technology is automatic (Limayem, Hirt & Cheung 2007) (Kim, Malhotra & Narasimhan 2005). The fourth
moderator, voluntariness of use, was removed from the revised UTAUT2 model because, in the consumer context of technology acceptance, customers can decide for themselves whether to adopt the new technology (Venkatesh et al., 2012).
2.3 Personal Innovativeness in the domain of IT
While the UTAUT2 model has seven determinants of the consumer’s behavioural intention to use a new technology, it fails to investigate individual traits, such as personal innovativeness. Previous studies, such as Goldsmith and Hofacker (1990), have shown that instead of
19 measuring a global view of innovativeness, a domain specific view of personal innovativeness will give retailers a more specific way to measure consumers’ behavioural intention to
purchase a certain product. From this perspective, Agarwal and Prasad conducted a study to define personal innovativeness in the domain of information technology (PIIT) (1998). Personal innovativeness can, therefore, be defined as ‘the willingness of an individual to try out any new information technology’ (Agarwal & Prasad, 1998) (shown in Figure 4).
2.3.1 Personal Innovativeness in the domain of IT as Moderator
The authors Agarwal and Prasad argue for the use of personal innovativeness in the domain of IT as a moderator on the relationship between the perceptions of a technology and the
behavioural intention of consumers (Rosen, 2005). Therefore, in the modified UTAUT2 model, Personal innovativeness is a moderator between all the factors that could influence consumers’ behavioural intention to use mobile AR retail applications (H7a-H7f). Moreover, Agarwal and Prasad only found a significant moderation effect between one determinant of the behavioural intention to use: compatibility. Therefore, it is of interest to retest their theory on the UTAUT2 model.
2.3.2 Personal Innovativeness in the domain of IT as a predictor of UTAUT2
In contrast with Agarwal and Prasad’s theory, in which they argue personal innovativeness in the domain of IT is a moderator, marketing researchers have found a direct link between personal innovativeness and the behavioural intention to use a new technology (Rosen, 2005). These researchers believe if highly innovative people are targeted first, they will eventually improve word-of-mouth advertising (Rosen, 2005). Limayem, Khalifa and Frini empirically tested this theory and found support for the direct positive link between PITT and purchase intentions (2000). As this thesis was interested in the determinants that increase the
behavioural intention to use mobile AR retail applications, the PIIT was also tested as a direct predictor of behavioural intention to use.
2.4 Conceptual Framework
After considering all available technology acceptance models, UTAUT2 was adopted for this research because this model is specially made to measure the consumer’s technology
acceptance. Furthermore, the UTAUT model explains as much as 70% of the variance in behavioural intention to use a technology (Venkatesh et al., 2012). Therefore, this model seems to be most appropriate for this study on consumers’ acceptance of mobile AR-based retail applications. To test the drivers for consumers’ acceptance of retail AR-based
applications, several hypotheses were formulated based on the currently proposed research model. The standard moderators of UTAUT2, age, gender and experience, were not included due to the time limit of this study. However, these moderators were used as control variables, which allowed for side analyses.
A new moderator, personal innovativeness in the domain of IT (Agarwal & Prasad, 1998), was tested. In addition, a direct link between this determinant and the behavioural intention to use mobile AR retail apps was tested.
21 The dependent variable use behaviour (US) was not researched because, due to the short time frame, it was not possible to measure the continuous use of the technology. As a result, behavioural intention was selected as the main dependent variable of this study.
The independent variable, price value (PC), was not included in this study because there are no monetary costs related to mobile AR-based apps used by consumers, as these apps are provided by the retailer and so free to download.
The independent variables of this study are as follows: performance expectancy, effort expectancy, social influence, facilitating conditions, hedonic motivation, habit and personal innovativeness in the domain of IT. These independent variables are proposed to have a positive direct effect on the behavioural intention to use mobile AR-based retail apps. In addition, personal innovativeness in the domain of IT is proposed to positively moderate the relationship between the first six independent variables and behavioural intention to use mobile AR-based retail app. In Figure 5, the conceptual model used in this research is presented.
2.5 Hypotheses
Hypotheses 1 through 6 (H1-H6) are designed to test which factors of the revised UTAUT2 model are validate for the consumers’ acceptance of mobile AR retail applications.
Performance Expectancy is defined as to what extent an individual believes the technology will help them to better completes their tasks (Venkatesh et al., 2003). Venkatesh et al. indicate the PE construct is the strongest predictor of behavioural intention to use a new technology (2003). As these mobile apps are developed to increase retailer’s m-commerce sales, the performance is likely to have a positive effect on consumers’ behavioural intention to use mobile AR-based retail apps. Therefore the following hypothesis is proposed:
H1: Performance expectancy will have a positive influence on behavioural intention to use mobile AR retail applications.
Effort Expectancy relates to what extent the individual believes the technology is easy to use (Venkatesh et al., 2003). In addition, EE is based on the perceived ease of use of the TAM model (Davis, 1989). Furthermore, this factor is expected to be important, as AR is new to mobile devices. In this study, it is therefore expected that if the consumer expects mobile AR-based retail apps to require little effort learn to use, this assumption could positively affect their behavioural intention to use these apps:
H2: Effort expectancy will have a positive influence on behavioural intention to use mobile AR retail applications.
Social influence relates to what extent others’ opinions influence an individual’s intention to use the technology, mostly in that individual’s own network of family and friends (Venkatesh
23 et al., 2003). Social influence is therefore expected to have a positive effect on the
behavioural intention to use mobile AR-based retail apps:
H3: Social influence will have a positive influence on behavioural intention to use mobile AR retail applications.
Facilitating conditions relates to what extent an individual believes he or she has the resources to make use of the technology (Venkatesh et al., 2003). A consumer needs to have a mobile device with software compatible with the AR function to make use of these AR-based retail apps. It is therefore expected that FC will positively influence the behavioural intention to use mobile AR-based retail apps:
H4: Facilitating conditions will have a positive influence on behavioural intention to use mobile AR retail applications.
Hedonic motivation is the extent to what an individual can enjoy the technology and derives pleasure from using it (Venkatesh et al., 2012). This determinant is important in consumer technology use (Brown & Venkatesh, 2005). As Rese et al. argued it is important to measure not only the functional utility of AR technology, but also the hedonic utility, as the use of new technologies is perceived as entertaining (2017). Therefore, it is expected that HM positively influences the behavioural intention to use mobile AR-based retail apps.
H5: Hedonic motivation will have a positive influence on behavioural intention to use mobile AR retail applications
Habit relates to the individual's previous use of the technology and the customer’s belief that the use of that particular technology is automatic to them (Limayem, Hirt & Cheung 2007) (Kim, Malhotra & Narasimhan 2005). Considering AR technology habitual is expected to
have a positive effect on the behavioural intention to use mobile AR-based retail apps. This is mainly because of the experience consumers obtained via other AR-based apps, such as Snapchat and Instagram filters.
H6: Habit will have a positive influence on behavioural intention to use mobile AR retail applications
Hypotheses 7a through 7f (H7a-H7f) measure whether personal innovativeness in the domain of IT is a positive moderator on the relationship between the factors and the behavioural intention to use mobile AR-based retail apps. Agarwal & Prasad’s paper illustrated that the more innovative consumers are, the more likely they are to develop positive perceptions that positively influence their behavioural intention to use a new technology (1998). As AR is in the begin phase of m-commerce, early adaptors are expected to positively moderates the relationship between the factors and the behavioral intention to use mobile AR-based retail apps. Therefore, this study hypothesized the following:
H7a: Personal innovativeness in the domain of IT positively moderates the
relationship between performance expectancy and behavioural intention to use, so this relationship is stronger for consumers with higher values of personal innovativeness.
H7b: Personal innovativeness in the domain of IT positively moderates the relationship between effort expectancy and behavioural intention to use, so this relationship is stronger for consumers with higher values of personal innovativeness.
H7c: Personal innovativeness in the domain of IT positively moderates the relationship between social influence and behavioural intention to use, so this
25 H7d: Personal innovativeness in the domain of IT positively moderates the
relationship between facilitating conditions and behavioural intention to use, so this relationship is stronger for consumers with higher values of personal innovativeness.
H7e: Personal innovativeness in the domain of IT positively moderates the relationship between hedonic motivation and behavioural intention to use, so this relationship is stronger for consumers with higher values of personal innovativeness.
H7f: Personal innovativeness in the domain of IT positively moderates the relationship between habit and behavioural intention to use, so this relationship is stronger for consumers with higher values of personal innovativeness.
Hypothesis 8 (H8) tests whether personal innovativeness in the domain of IT is a determinant of the behavioural intention to use mobile AR retail applications, as this study is focused on the determinants. According to Rosen (2005), marketing researchers found a direct positive link between PIIT and behavioural intention to use, and therefore, this study hypothesized the following.
H8: Personal innovativeness in the domain of IT will have a positive influence on behavioural intention to use mobile AR retail applications
3. Data & Method
This quantitative research is an explanatory study that obtained its data via a cross-sectional survey design, as this is the most appropriate when testing a particular phenomenon at a particular time. Four different surveys were distributed. Each focused on a specific retail segment: Beauty and Cosmetics, Food, Home Décor and Fashion Items. The survey data was used to test the hypotheses and generalize research and managerial implications. The online survey consisted of multiple 5-point Likert-scale questions and the obtained data was analysed in SPSS. The survey was translated into Dutch to lower language barriers to encourage respondents to complete the survey. In order to verify the translation the survey was checked by individuals who are proficient in both English and Dutch (Appendix 1).
3.1 Methodology
The online survey asked the respondents to complete demographical questions about their gender (nominal variable), age (ratio variable) and educational background (ordinal variable). The other constructs from the questionnaire were adopted from the original UTAUT2 paper from Venkatesh et al. (2012) and were therefore measured using Venkatesh et al.’s (2012) scale. A validated Likert Scale was used on a 5-point scale ranging from ‘strongly disagree’ (1) to ‘strongly agree’ (5) at the interval level. Some scales were shortened because using all items would make the survey too long, risking a lower response rate. Personal innovativeness was measured with three items on Agarwal and Prasad’s 5-point Likert-scale of (1998). All the scales and corresponding items are shown in Table 2.
27 Table 2. UTAUT2 survey items (Venkatesh et al., 2012)
Scale Items
Performance Expectancy
PE1. I find these kinds of Augmented Reality Retail Applications useful in my daily life
PE2. Using these kinds of Augmented Reality Retail Applications increases my chances to make better purchase decisions
PE3. Using these kinds of Augmented Reality Retail Applications helps me to make purchase decisions quicker
Effort Expectancy
EE1. Learning how to use such Augmented Reality Application is easy for me EE2. My interaction with Augmented Reality Applications is clear and understandable
EE3. I find these Augmented Reality Applications easy to use (or perception) EE4. It is easy for me to become skilful at using these Augmented Reality Applications
Social Influence
SI1. People who are important to me think that I should use Augmented Reality Retail Applications
SI2. People who influence my behaviour think that I should use Augmented Reality Retail Applications
SI3. People whose opinions that I value prefer that I use Augmented Reality Applications
Facilitating Conditions
FC1. I have a mobile phone (IOS 11 or Samsung S9) where I can use these Augmented Reality Applications on
FC2. I have the knowledge necessary to download and use Augmented Reality Applications
FC3. I can get help from others when I have difficulties using Augmented Reality Applications
Hedonic Motivation
HM1. Using Augmented Reality Applications is fun HM2. Using Augmented Reality Applications is enjoyable HM3. Using Augmented Reality Applications is very entertaining
Habit H1. The use of Augmented Reality Applications has become a habit for me H2. I am addicted to using Augmented Reality Applications
H3. I must use Augmented Reality Applications
H4. Using Augmented Reality Applications has become natural to me Personal
Innovativeness
PIIT1. If I heard about a new information technology, I would look for ways to experiment with it
PIIT2. Among my peers, I am usually the first to try out new information technologies
PIIT3. I like to experiment with new information technologies Behavioural
Intention BI1. I intend to continue using these kinds of Augmented Reality Applications in the future BI2. I will always try to use these kinds of Augmented Reality Applications in my daily life
BI3. I plan to continue to use these kinds of Augmented Reality Applications frequently
3.2 Data collection and sample
The survey was distributed online via personal social media channels and public channels, such as forums and websites, from 19 April to 7 May 2018. The total number of respondents
to the survey was 254. However, 24 respondents did not complete the whole survey, and therefore, these incomplete surveys were removed from the sample.
From the remaining 230 respondents, 198 indicated that they had used mobile AR applications before. Since this study wanted to test the acceptance of mobile AR apps in retail, only the data of the respondents who knew how this AR technology works was used.
The sample was checked for normality; all histograms of frequencies were bell shaped. However, the Kolmogorov-Smirnov test failed, indicating, the null hypothesis of normal distribution should be rejected. However, almost all the QQ plots (appendix 2) were on the diagonal, which means they are normally distributed. According to Field, for a large sample size (±200), it is more important to look at the shape of the distribution and the statistical value of the skewness and kurtosis (2009). The skewness and kurtosis values for this research were all between -1 and 1, except for the variable HM. Therefore, the outliers were examined and trimming the data reduced the impact of bias. Using the boxplot test, 11 extreme outliers were identified and excluded for further analyses. Furthermore, the Z-scores were checked for values above 3 and below -3; these outliers were also excluded from further analyses. The remaining 179 respondents were used as the sample size for this study. Again, a normality check was performed, and all skewness and kurtosis values were acceptable values between -1 and 1 (Field, 2009), shown in appendix 3. Of the variables, BI, PE, FC and PIIT had negative skewness, indicating a pile-up of scores on the right of the distribution, while EE, SI, HM, and H had positive skewness, indicating a pile-up to the left of the distribution. The positive values of kurtosis of BI, PE, FC and HM indicate a pointy and heavy-tailed distribution. The negative values of kurtosis of EE, SI, H and PIIT indicate a flat and light-tailed distribution (Field, 2009).
29 Of respondents, 64.8% were female, while 35.2% were male. The sample was skewed to more highly educated people, as displayed in Table 3, where 79.3% of the people had a high educational level: applied sciences, university bachelor’s, master’s or PhD. The education groups were split into low, middle and high educated according to the national guidelines of Centraal Bureau Statistiek (CBS, 2016).
Table 3. Educational Level
Frequency Percent Valid Percent
Cumulative Percent
Valid Low Educational Level 2 1.1 1.1 1.1
Middle Educational Level 35 19.6 19.6 20.7
High Educational Level 142 79.3 79.3 100.0
Total 179 100.0 100.0
As Table 4 illustrates, the sample mainly represents a young population between the ages of 16 and 25 years old (81.6%). According Strauss and Howe’s (2007) book, Millennials Go To College, the age groups can be divided into the Baby Boomer Generation (1943-1960), Generation X (1961-1981), Millennials (1982-2003) and Generation Z (2004 and later). Thus, using Strauss and Howe’s guidelines, 93.9% of the respondents belong to the Millennial generation.
Table 4. Age Groups
Frequency Percent Valid Percent
Cumulative Percent Valid 16-25 146 81.6 81.6 81.6 26-35 22 12.3 12.3 93.9 36-45 10 5.6 5.6 99.4 46-55 1 0.6 0.6 100.0 Total 179 100.0 100.0
4. Results
In this chapter, the results of the study’s analyses are described. First, the preliminary steps were completed. Second, the reliability and validity tests are presented. Third, the results of the regression, stepwise regression, process and one-way ANOVA are described. Last, the hypotheses’ testing is described.
4.1 Preliminary steps
First, to enable the use of gender as a control variable, (1) male and (2) female were recoded into male = 0 and female = 1. Furthermore, the categorical variable education was changed into dichotomous variables, also called dummy coding, as this categorical variable had more than two categories. This step was necessary due to the regression analysis that was
performed in later stage, in which education was used as a predictor. None of the variables has items that were counter-indicative, so none had to be recoded.
4.2 Reliability and Validity
A reliability analysis was executed on all factors. Looking at the Cronbach’s alpha value is commonly used to test the reliability of a scale. Examining the correlation between the different items tests the reliability. A Cronbach's alpha value above .7 is an acceptable value; values substantially lower (α < .7) indicate unreliability (Field, 2009). The PE scale had high reliability, with a Cronbach’s alpha of .797. The corrected item-total correlations indicate that all the items had a good correlation with the total score of the scale, as all scored above .030. Also, none of the items would substantially affect reliability if one were deleted, by looking if Cronbach’s alpha if item deleted was substantially different from the Cronbach’s alpha value tests this.
The following factor scales were also checked for reliability. The EE scale had high reliability, with a Cronbach’s alpha of .798. For the scale of SI, a high reliability was present
31 reliability was high, with a Cronbach’s α of .770. For personal innovativeness, Cronbach's α was .911, and for behavioural intention, the α was .760. However, the reliability analysis indicated a problem with the scale of FC. The Cronbach’s alpha for this factor was .461, which is below the minimum of .7. Even when one of the items was deleted, the Cronbach’s alpha remained too low. Since the assumption is that there must be at least three items, the decision has made to reject this factor for further analyses. All outcomes are presented in Table 5.
Table 5 Reliability, Cronbach's alpha
Cronbach's alpha Performance expectancy .797 Effort expectancy .798 Social influence .946 Facilitating conditions .461 Hedonic motivation .844 Habit .770 Personal innovativeness .911 Behavioural intention .760
4.2.1 Principal Components Analysis
A principal components analysis (PCA) was performed on the scales of the independent variables to examine similarities between the variables and to check whether this study could reduce the set of variables into a number of factors. Since the dependent variable was known, there was no need to use this variable in the factor analysis (Field, 2009). The Kaiser-Meyer-Olkin (KMO) measure verified the sampling adequacy for the analysis (KMO = .777), which is above the minimum (> |.60|) and Bartlett’s Test of Sphericity X2 was significant (< .001): (253) = 2232.98, shown in Table 6. This indicates that correlations between the items were
sufficiently large for the PCA analysis. Seven components had eigenvalues over Kaiser’s criterion of 1, and in combination, these components explained 73.39% of the variance.
Another method to check whether an eigenvalue is large enough to represent a factor is by examining the scree plot, shown in Appendix 4. According to Cattell (1966b), the cut-off point for selecting factors should be at the point of inflexion of this curve, also known as the point where the slope of the line changes dramatically. The scree plot examined a point of inflexion after the seventh factor. Therefore, seven factors were retained and rotated with an oblimin with Kaiser normalization rotation, a method of oblique rotation (Field, 2009). With oblique rotation, the factors are allowed to correlate.
Table 7 shows the factor loadings after the oblimin rotation. The items that cluster on the same factor suggest that factor 1 represents HM, factor 2 represents SI, factor 3 represents PIIT, factor 4 represents PE, factor 5 represents EE, factor 6 represents H and factor 7
represents FC. The second item of facilitation conditions indicates high cross loading on the factor of HM. The same is true for the last item of FC, which shows high cross-loadings on the factor of HM as well. Both results could be due to the content of the items. However, as mentioned in the reliability analysis, the factor FC was rejected from further analyses due to a Cronbach’s alpha of less than .7. According to Field, the pattern matrix shows evidence for convergent validity if all the factor loadings have a score great than .4 (2009). If FC is disregarded, all the loadings had scores greater than .6, which indicates high validity. Table 6. KMO and Bartlett's Test
Kaiser-Meyer-Olkin Measure of Sampling Adequacy. .777
Bartlett's Test of Sphericity Approx. Chi-Square 2232.984
df 253
33 Table 7. Factor Loading PCA
Pattern Matrix
Rotated Factor Loadings
HM SI PIIT PE EE H FC
Q11-1 I find these kinds of Augmented Reality Retail Applications useful in my daily life
.129 .014 -.081 -.788 -.083 .106 -.143
Q11-2 Using these kinds of Augmented Reality Retail Applications increases my chances to make better purchase decisions
.007 -.032 .116 -.848 .080 -.027 .043
Q11-3 Using these kinds of Augmented Reality Retail Applications helps me to make purchase decisions quicker
-.018 -.117 -.053 -.804 -.003 .003 .198
Q12-1 Learning how to use such Augmented Reality Application is easy for me.
.055 .063 .084 -.145 -.549 -.082 .260
Q12-2 My interaction with Augmented Reality Applications is clear and understandable.
-.042 .037 -.023 -.008 -.889 -.018 -.110
Q12-3 I find these Augmented Reality Applications easy to use (or perception)
.000 -.026 -.073 .021 -.831 .195 .032 Q12-4 It is easy for me to become
skilful at using these Augmented Reality Applications.
.047 -.120 .109 .056 -.784 -.082 -.063
Q13-1 People who are important to me think that I should use Augmented Reality Retail Applications
-.025 -.929 -.006 -.047 -.054 -.010 -.045
Q13-2 People who influence my behaviour think that I should use Augmented Reality Retail Applications
.043 -.929 .056 -.022 -.013 .000 .024
Q13-3 People whose opinions that I value prefer that I use Augmented Reality Applications
.011 -.928 .009 -.032 .012 .039 .007
Q14-1 I have a mobile phone (IOS 11 or Samsung S9) where I can use these Augmented Reality Applications on
-.079 .039 .083 -.100 .044 .043 .896
Q14-2 I have the knowledge necessary to download and use Augmented Reality Applications
.339 -.041 .051 .042 -.329 -.052 .419
Q14 -3 I can get help from others when I have difficulties using Augmented Reality Applications.
.637 -.097 -.113 .157 .115 .089 .222
Q15-1 Using Augmented Reality Applications is fun.
.801 -.022 .059 -.104 -.086 .012 -.120 Q15-2 Using Augmented Reality
Applications is enjoyable.
.710 -.129 .040 -.124 -.096 .040 -.089 Q15-3 Using Augmented Reality
Applications is very entertaining.
.890 .124 .091 -.080 -.002 -.063 -.084 Q16-1 The use of Augmented Reality
Applications has become a habit for me.
.086 -.069 .030 .025 -.026 .792 .111
Q16-2 I am addicted to using
Q16-3 I must use Augmented Reality
Applications -.039 .003 .074 -.193 .046 .630 -.178
Q16-4 Using Augmented Reality Applications has become natural to me.
.068 .127 .061 .020 -.127 .798 .164
Q17-1 If I heard about a new
information technology, I would look for ways to experiment with it.
.033 -.020 .912 -.002 -.014 .032 -.025
Q17-2 Among my peers, I am usually the first to try out new information technologies.
.041 .005 .906 .068 .012 .093 .023
Q17-3 I like to experiment with new information technologies.
-.043 -.033 .917 -.012 .014 -.031 .044 Extraction method: PCA
Rotation method: oblimin with Kaiser normalization.
4.2.2 Convergent validity
As this study wanted to assess the total validity of the model, the convergent validity is inspected. The convergent validity is a measure of the amount of variance captured by a construct in relation to the amount of variance due to measurement error (Fornell & Larcker, 1981). The convergent validity is measured using the values of the average variance extracted (AVE). According to the criterion of Fornell and Larcker, a minimum AVE value of .5 is needed (1981). For this study, all factors met this criterion. The AVE for PE was .662. For EE, the AVE was .599; for SI, it was .862; for HM, it was .646; for H, it was .550 and for PIIT, it was .831. These outcomes are presented in Table 8.
4.2.3 Correlation Matrix
Before the correlation matrix was performed, means were first computed. The items per factor were computed into one and standardized. Next, a bivariate Pearson correlation was
performed. Results displayed in Table 8. The correlation matrix indicated no Pearson’s correlation coefficients were above .9. Therefore, this study assumed that multicollinearity was not a problem (Field, 2009). There was a statistically significant (p < 0.01) positive linear relationship between PE and behavioural intention (.520). However, the correlation
35 coefficient did not indicate in which direction the causality operated between these two
variables. This positive relationship also applied to the relationship between PIIT and BI (.487) and between HM and BI (.395). Furthermore, there were statistically significant (p < .01) positive linear relationships between EE and BI (.231), between SI and BI (.356) and between H and BI (.312).
Variables M SD 1 2 3 4 5 6 7 8 9 10 AVE .662 .599 .862 .646 .550 .831 1.BI 3.4 .649 (.760) 2.Age 24.41 5.127 .138 3.Gender 0.65 .479 -.205** -.113 4.Education 6.54 1.435 .136 .379** .039 5.PE 3.53 .729 .520** -.035 -.027 .016 (.797) 6.EE 4.13 .484 .231** .028 -.082 .160* .209** (.798) 7.SI 2.33 .861 .356** .084 -.088 .022 .325** .109 (.946) 8.HM 3.99 .547 .395** .021 -.058 .101 .319** .430** .264** (.844) 9.H 2.18 .781 .312** -.013 -.090 -.145 .270** .106 .386** .205** (.770) 10.PIIT 3.12 1.069 .487** .049 -.299** .114 .223** .268** .154* .249** .305** (.911)
**. Correlation significant at the 0.01 level (two-tailed). *. Correlation significant at the 0.05 level (two-tailed).
37 4.3 Analyses
First, the results of the multiple linear regression analyses will be discussed to find the
maximum explained variance model. Thereafter, the stepwise linear regression, used to find a model with a maximum of statistically significant variables, is presented. Third, the
moderation interaction is explained by using process, and afterward, the one-way ANOVA is discussed.
4.3.1 Hierarchical Regression Model
For the first model only the control variables, age, gender and education, were measured as predictors of behavioural intention to use, and the model explained 5.1% of the variance. The model was statistically significant: F (3,175) = 4.196, p < .01. These results indicate most variance of the variable behavioural intention to use is explained by other factors.
In model 2, the predictors PE, EE, SI, HM, H and PIIT were added, and the total variance explained by the model as a whole was 45.3% (F (9,169) = 17.387), which was significant (p < .001). In the second model, three out of nine predictor variables were significantly correlated, with PE recording the highest standardized beta coefficient (β = .355, p < .001). Thus, if PE increases by one, the behavioural intention to use mobile AR retail applications will increase by .355. The second largest predictor correlated with BI was PIIT with the following standardized beta coefficient: β = .298, p < .001. The third predictor correlated with BI was HM (β = .168, p < .05). All three paths mentioned were positively related to the behavioural intention to use mobile AR retail apps. The three paths were consistent with H1, H5 and H8, which were associated with the intention to use mobile AR apps as a retail tool, such that a more positive PE, HM and PIIT, respectively, lead to a greater BI. The results of the hierarchical regression model are shown in Table 9.
Hypotheses H2, H3 and H6 were non-significant (p > .05), so that a higher level of EE (H2, β = -.032, p = .615), SI (H3, β = .119, p = .062) and H (H6, β = .051, p = .429) were not
associated with a significantly higher or lower level of behavioural intention to use mobile AR apps in retail.
In the third model, the interaction terms were also included, and the total variance explained by the model as a whole was 45.8% (F (14,164) = 11,736), which was not significant (p > .05). Therefore, the moderation effect of personal innovativeness does not explain a higher variance. The interaction term was also analysed in process (Hayes, 2013), which is elaborated in 5.3.3 in the stepwise regression, elaborated in section 5.3.2.
Furthermore, only gender was statistically significant in model 1, which measured only the control variables as explanatory variables. In the other two models, none of the control variables was statistically significant. This result means that age, gender and education level do not influence the results of the predictors influencing the behavioural intention to use mobile AR retail apps.
39 Table 8. Hierarchical Regression Model of behavioural Intention
Behavioural Intention
Model 1 Model 2 Model 3
Explanatory variables
B S.E. Beta B S.E. Beta B S.E. Beta
Control Variables Age .009 .010 .071 .012 .008 .094 .013 .008 .104 Gender -.273 .100 -.201** -.103 .080 -.076 -.141 .082 -.104 Education .053 .036 .117 .026 .028 .057 .018 .029 .040 Predictor Variables PE .316 .055 .355*** .318 .055 .356*** EE -.043 .085 -.032 -.025 .085 -.019 SI .090 .048 .119 .111 .049 .148* HM .199 .077 .168* .191 .079 .161* H .042 .054 .051 .039 .056 .047 PIIT .181 .039 .298*** .171 .042 .282*** Moderator Variables PE x PIIT .079 .051 .102 EE x PIIT -.110 .076 -.101 SI x PIIT -.081 .043 -.126 HM x PIIT .064 .073 .067 H x PIIT .011 .051 .014 F-value 4.196** 17.387*** 11.736*** df (df1, df2) (3,175) (9,169) (14,164) R2 .067** .481*** .500 dR2 .051 .453 .458 ***p < .001 **p < .01 *p < .01
4.3.2 Stepwise Linear Regression Model
A stepwise linear regression model was used to perform a regression on multiple variables and simultaneously remove the insignificant variables. In this method, it is not possible to lose a significant variable and not possible to have an insignificant variable. Therefore, this model results in the best explanatory variables. In this stepwise linear regression model, the control variables and the other predictors, including the interaction terms, are measured. As presented in Table 10, five different models (in each model a predictor was added) were found
Table 9 stepwise regressions, model summary
Model Summaryf
Model R R Square Adjusted R Square R Square Change F Change df1 df2 Sig. F Change 1a .52 .271 .267 .271 65.714 1 177 .000 2b .639 .409 .402 .138 41.049 1 176 .000 3c .663 .439 .429 .030 9.447 1 175 .002 4d .677 .458 .446 .019 6.201 1 174 .014 5e .687 .472 .457 .014 4.624 1 173 .033 a. Predictors: (Constant), PE b. Predictors: (Constant), PE, PIIT c. Predictors: (Constant), PE, PIIT, HM d. Predictors: (Constant), PE, PIIT, HM, SI e. Predictors: (Constant), PE, PIIT, HM, SI, Age f. Dependent Variable: BI
The five predictors in model 5 explained 45.7% of the total variance (F (1,173) = 4,624) and were significant (p < .05). Performance expectancy was the greatest predictor of the behavioural intention to use mobile AR retail apps and explained 26.7% of the variance. The second largest predictor for behavioural intention to use mobile augmented reality retail apps was personal innovativeness, which explained another 13.8% of the variance. The third predictor was HM, which was much lower than the other two, explaining only an additional 3% of the variance. Last, two additional small predictors were found statistically significant (p < .05), SI and age, which respectively explained 1.9% and 1.4% of the variance. Therefore, five predictors were significant: PE, personal innovativeness, HM, SI and age. Four of these predictors are consistent with H1, H3, H5 and H8.
Table 11 presents the coefficients in model 5. In this model, PE had the highest standardized beta coefficient (β = .355, p < .001). Therefore, if PE increases by one, the behavioural intention to use mobile AR retail apps will increase by .355. If PIIT increases by one, the behavioural intention to use mobile AR retail apps will increase by .332. For HM, the increase would be lower, as the behavioural intention to use would increase by .161. If the
41 fourth predictor, SI, increases by one, the behavioural intention to use will increase by .137. Last, the predictor age had a direct positive relationship with the behavioural intention to use; if someone’s age increases by one, the BI will increase by .120. None of the interaction terms were added to the model, so none of these interaction terms were statistically significant.
Table 10. Stepwise regression, coefficients Coefficients Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 5 (Constant) .276 .329 0.841 .402 PE .316 .054 .355 5.798 .000 PIIT .202 .035 .332 5.731 .000 HM .191 .071 .161 2.676 .008 SI .104 .045 .137 2.303 .022 Age .015 .007 .120 2.150 .033 Dependent Variable: BI
Additionally, three case numbers had residual sizes that exceeded 3. These cases were not excluded, as they are not the result of a mistake during data entry. Furthermore, the Cook’s distance value was below 1. In addition, the three case numbers that exceeded 3, only accounts for 1.67% of the total sample size; therefore, the outliers have a relatively small impact on the model.
The histogram (Figure 6) of the residuals suggests that they are close to being
normally distributed. The scatterplot in Appendix 5 indicates the residuals are not distributed in any pattern with the predicted values. Therefore, the model does not violate the assumption of homoscedasticity.
Figure 6. Histogram of model 5
4.3.2 Moderation effect
After the regression analyses, the simple moderation effect of personal innovativeness in the domain of IT was analysed by the process analysis of Hayes (2013). Process is a macro tool for SPSS, to check whether personal innovativeness positively moderated the relationship between the predictor of the model and the outcome. For this reason, the p-value of the interaction term was checked for each interaction term in the model to determine whether these interaction terms were significant to the model. The interaction terms were consistent with hypotheses 7a through 7f.
The first interaction term PE x PIIT was not statistically significant (p = .2208, so p > .05) (shown in Table 12). Thus, the effect of PE on the behavioural intention to use mobile AR apps in the future does not depend on someone’s personal innovativeness in the domain of IT. Therefore, H7a was rejected.
43 Table 12. Process moderation effect of personal innovativeness on PE
Coefficient SE t p
Intercept -.0039 .0547 -.0718 .9428
PE (X) .4241 .0579 7.3299 .0000
PIIT (W) .3566 .0557 6.3972 .0000
PE*PIIT (XW) .0655 .0533 1.2288 .2208
The second interaction term EE x PIIT was also not statistically significant (p = .8341, so p > .05) (shown in Table 13). Thus, the effect of EE on the behavioural intention to use mobile AR apps in the future does not depend on someone’s personal innovativeness in the domain of IT. Therefore, H7b was rejected.
Table 13. Process moderation effect of personal innovativeness on EE
Coefficient SE t p
Intercept .0250 .0626 .4000 .6896
EE (X) .1095 .0673 1.6282 .1053
PIIT (W) .4229 .0649 6.5114 .0000
EE*PIIT (XW) -.0121 .0578 -.2097 .8341
The third interaction term SI x PIIT was not statistically significant (p = .3634, so p > .05) (shown in Table 14). Thus, the effect of SI on the behavioural intention to use mobile AR apps in the future does not depend on someone’s personal innovativeness in the domain of IT. Therefore, H7c was rejected.
Table 14. Process moderation effect of personal innovativeness on SI
Coefficient SE t p
Intercept .0301 .0585 .5141 .6078
SI (X) .2919 .0628 4.6484 .0000
PIIT (W) .4056 .0595 6.8149 .0000
The fourth interaction term HM x PIIT was not statistically significant (p = .9670, so p > .05) (shown in Table 15). Thus, the effect of HM on the behavioural intention to use mobile AR apps in the future does not depend on someone’s personal innovativeness in the domain of IT. Therefore, H7e was rejected.
Table 15. Process moderation effect of personal innovativeness on HM
Coefficient SE t p
Intercept .0210 .0595 .3535 .7242
HM (X) .2772 .0617 4.4930 .0000
PIIT (W) .3804 .0622 6.1197 .0000
HM*PIIT (XW) -.0023 .0554 -.0414 .9670
The last interaction term H x PIIT was not statistically significant (p = .0983, so p > .05) (shown in Table 16). Thus, the effect of H on the behavioural intention to use mobile AR apps in the future does not depend on someone’s personal innovativeness in the domain of IT. Therefore, H7f was rejected.
Table 16. Process moderation effect of personal innovativeness on H
Coefficient SE t P
Intercept -.0085 .0627 -.1349 .8929
H (X) .1344 .0689 1.9511 .0526
PIIT (W) .4133 .0644 6.4140 .0000
H*PIIT (XW) .1080 .0650 1.6621 .0983
According to the process analyses of Hayes, there is no moderation effect; meaning hypotheses H7a through H7f were all rejected (2013).
4.3.3 One-way ANOVA
A one-way ANOVA was used to discover whether there were any differences in acceptance among the four retail sectors. The results, in Table 17, have shown that there was no
45 apps in the future: F (3, 175) = 1.766, p = .155. Therefore, there was no significant difference in acceptance among the four retail sectors (p > .05).
Table 17. One-way ANOVA retail segments
As Table 17 illustrates, the four retail sectors each have different means for the behavioural intention to use mobile AR retail apps; therefore, there is indeed a difference in behavioural intention, although these differences are not significant. This result could be due to the low sample size for each retail sector, which is between 37 and 51 respondents.
The means plot in Appendix 4 shows that the Home sector (group 1) had a greater behavioural intention than the other three groups to use these mobile AR retail apps in the near future. However, for the beauty sector (group 3), the behavioural intention to use was lower than for the food sector and for the fashion sector.
4.3.4 Additional side analyses
As mentioned before, this study measured three main control variables: age, gender and education. For all these control variables a one-way ANOVA was performed. The results are described below.
Gender
As Table 18 illustrates, there was a statistically significant effect of gender on the behavioural intention to use mobile AR retail apps in the future: F (1,177) = 7.759, p < .01. Therefore, males had a higher behavioural intention than females to use mobile AR retail apps in the future (Appendix 6).
SS DF MS F Sig. Sector Mean SD N
Sector 2.206 3 .735 1.766 .155 1.Home 3.54 .54 51
Error 72.855 175 .416 2.Food 3.38 .78 42
Total 75.060 178 3. Beauty 3.22 .58 37
4. Fashion 3.40 .66 49
