Master thesis Are Network Externalities the Key for Consumers to Adopt Platform Technology?

(1)

0 THE EFFECT OF NETWORK EXTERNALITIES ON PLATFORM ADOPTION

Master thesis

Are Network Externalities the Key for Consumers to Adopt Platform Technology?

A Quantitative Study on the Moderating Role of Network Externalities on the Potential Consumers’ Buying Intentions of Platform Technology

Student: Anne Antonia Anna Peters Student number: 12392014

University of Amsterdam / Amsterdam Business School Executive Programme in Management Studies – Digital Track

Track: Digital business Supervisor: M.T. Ramezan Zadeh

Date: 29-01-2021 Version: Final

EBEC Approval number: EC 20201002021050

(2)

1

Statement of Originality

This document is written by student Anne Antonia Anna Peters 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 content.

Signature:

Anne Antonia Anna Peters

(3)

2

Abstract

A platform strategy is becoming more and more important for companies in the industry of Internet of Things. This study examines the effect of network externalities on the relationship between perceived ease of use, perceived usefulness and behavioural intention to buy a (potential) platform complementor. Network externalities are one of the platform-related variables influencing the growth of platform technologies. This study suggests a theoretical framework that perceived ease of use positively influences the intention to buy a (potential) platform complementor and when network externalities are high this relationship will be strengthened. This study also examines a moderated mediation model. Perceived ease of use is also positively influencing the intention to buy a (potential) platform complementor through perceived usefulness. This mediation effect occurs by the fact the easier a product is to use, the usefulness will increase. This study proposes that network externalities as moderator will strengthen this mediated relationship.

206 respondents in the Netherlands answered online questionnaires focused on the research variables. My analysis reports that the direct relationship and the mediation effect are confirmed. Perceived ease of use is positively related towards behavioural intention to buy (β

= 0.17, p <0.01). Next to this, perceived ease of use is positively related towards behavioural intention to buy through perceived usefulness (β = 0.36, p <0.05). Furthermore, unexpected not supported results on the moderation effect are found. Because of that, discussion also focuses on the unconfirmed effect of network externalities as moderator in platform adoption.

In this research, I studied the case of smart home technology, which provides insight into the consumers’ intention to buy this technology. Moreover, this study suggests limitations and suggestions for future research.

Key words: platform management, network externalities, behavioural intention to buy, smart home technology, robot vacuum cleaner.

(6)

5

1. Introduction

Technology companies emerge all over the world and satisfy consumers with their services and products. These companies grow due to various factors, one of which is network externalities (Gallaugher, 2013). Network externalities are one of the platform-related variables that have an effect on the adoption of technology by consumers, especially of platform technology (Hsu & Lin, 2016; Lin & Bhattacherjee, 2008; Shin, Park, & Lee, 2018;

Xiao, Fu, & Liu, 2018). Katz & Shapiro (1985) define network externalities as: “The value or effect that consumers obtain by a bigger scale of users or more products or services”. This study focuses on network externalities: perceived network size and perceived

complementarity. Perceived network size is the consumers’ perception of the number of other consumers, which are using or will use the same service (Hsu & Lin, 2016), while perceived complementarity is about the range of complementary supporting services and applications on a platform that the consumer acquires (Park, Kwak, Lee, & Ahn, 2018).

Platform technology is defined by Gallaugher (2013) as: “Products and services that allow for the development and integration of software products and other complementary goods, effectively creating an ecosystem of value-added offerings. Windows, iOS, the Kindle, and the standards that allow users to create Facebook apps are all platforms”. Another

example of a platform technology with many users and which benefits from the network externalities is Amazon. The consumer buys books, streams videos (Prime Video), and buys devices (Alexa) from the same platform. Moreover, Amazon funds developers to provide new apps and technologies. The voice-activated Echo products is an example of this.

On a smaller scale, smart home technology is also a platform technology (Ehrenhard, Kijl, &

Nieuwenhuis, 2014), because it connects electrical appliances and services via a

communication network in a resident’s home to control, monitor, and access their residence remotely (Nikou, 2019; Kim, Park, & Choi, 2017).

(7)

6

1.1. Research problem

Consumers have various motives and barriers to buy an innovative product of the industry of Internet of Things (IoT). Therefore, research has conducted on the consumers’ intention of buying technologies and innovation (Ajzen, 1985; Davis, 1989). Market research shows an expected annual growth of 12% for the smart home technology market (Hong, Nam, & Kim, 2020). Although this trend seems to have a positive effect, a difference of adoption behaviour between early consumers and the mass market appears. This difference means that only a small group adopts the technology, which results in a more flattened growth than expected.

The reason for this discrepancy is not researched extensively (Shin et al., 2018). One of the reasons could be that end users are not yet convinced of the benefits of smart home

technology benefits (Ehrenhard et al., 2014). Another reason could be the quality of the products connected to the platform. However, the effect between product-related variables and platform-related variables is not extensively researched yet (Park et al., 2018). Examples of product-related variables are price, design, and functionality. The behaviour of consumers to adopt platform technology is partially unknown and should be investigated extensively. Via this way, manufacturers and sellers can use this information to put their platforms with

corresponding products more improved in the market (Kim et al., 2017; Kuebel, Hanner, &

Zarnekow, 2015; Nikou, 2019;).

1.2. Contemporary literature

Many applications and services are developed and accessible for consumers to integrate into smart home technology to facilitate the consumer (Hubert, Blut, Borck, Zhang, Koch &

Riedl, 2019; Kim et al., 2017; Nikou, 2019). For example, a fridge connected to the internet can order new milk when it is empty or a doorbell at the front door that is remotely accessible by the consumer. Contemporary research has given insights into these benefits and the

(8)

7 adoption of these products and their platforms (Shin et al., 2018). In many studies network externalities are giving this positive effect as benefit to consumers which strengthens the adoption of platforms (Hsu & Lin, 2016; Lin & Bhattacherjee, 2008; Zhao & Lu, 2012).

In particular, platform management is stated as a market barrier (Ehrenhard et al., 2014).

Platform management is applied by manufacturers and suppliers as an important strategy.

Ehrenhard (2014) states that it is crucial to integrate systems and make standards within the platform, this results in high interoperability, a benefit for the consumer. Interoperability, high standardisation, and integrated systems can be seen as benefits. These benefits result in a higher number of users and an increase of complementarity products integrating with the platform (Park et al., 2018).

1.3. Research gap

Literature that contradicts the convenience of smart home technology shows the non-effect of platform variables on the adoption of smart home technology (Lin & Bhattacherjee, 2008).

An example of a non-effect is congestion (Katz & Shapiro, 1985).This happens when too many users block the road which will slow down the internet. Additionally, research of Goldenberg, Libai & Muller (2010) show that network externalities result in chilling effects and slow down growth. This happens when the mass market waits for adopting smart home technology before early adopters due to low network externalities. In other words, less

complementary products and less users will negatively influences the adoption of smart home technology.

These network externalities are investigated as direct and indirect effects (Lin &

Bhattacherjee, 2008; Park et al., 2018). For example, the direct effect of perceived network size is the perceived benefits of IT usage and perceived ease of use (Park et al., 2018). This means that a higher network size gain a higher ease of use and more benefits. The indirect effect can be found in the positive effect of perceived complementarity on intention to buy,

(9)

8 mediated by perceived ease of use (Hsu & Lin, 2016). By including and excluding the

network externalities as factors, it is possible to gain knowledge on the positive and negative sides of platform adoption theories. Not only the direct and indirect effects are investigated, but also the moderator effects. The moderator according to Sharma, Durand and Gur-Arie (1981) is “One which systematically modifies either the form and/or strength of the

relationship between a predictor and criterion variable”. An example of moderation effect is the influence of gender and age on the relationship between perceived ease of use and the intention to buy smart home technology. This moderator weakens or strengthens the relationship between two variables (Abubakar & Ahmad, 2013). Propositions are made to extend research on the adoption of technology platforms with moderators (Baptista &

Oliveira, 2015; Guldentops, 2018; Park et al., 2018).

1.4. Research purpose and contribution

As mentioned above, a platform can consist of many services and products. Therefore, it is interesting to investigate possible products that can connect to this platform (Park et al., 2018). In this research, it will be referred as potential platform complementor. The use of the word potential in this is, because of the reason that services and products are also able to be used stand-alone.

Testing network externalities as moderators on this potential platform complementor creates explanatory power for the relationship of the independent variable to the intention of potential consumers to buy a platform complementor. The network externality is a contextual factor for platforms, by adding this factor the explanatory power of adoption theories will be stronger (Sun & Zhang, 2006). Trust is an example of the effect on the relationship between perceived ease of use and intention to buy. The higher the trust, the stronger the relationship between perceived ease of use and intention to buy. Testing the effect of moderators gains more insight into consumer behaviour (Guldentops, 2018). Also, it is stated that it is useful to

(10)

9 examine these and other possible explanations for factors that might moderate the relationship between ease of use and usage (Adams, Nelson, & Todd, 1992). Perceived complementarity for example affects the relationship between perceived ease of use and behavioural intention of buying smart home technology in a positive way (Park et al., 2018). This means the easier a system is to use, the higher the intention of buying and this will be strengthened by a high perceived complementarity. Moreover, network size shows a significant positive influence on the adoption of technology (Lin & Bhattacherjee, 2008; Lin & Lu, 2011). This means when the number of users grow, the relationship between perceived ease of use and intention of buying will strengthen. Next to this, former research doubts the effect of network externalities negatively (Goldenberg, et al., 2010). As the urge for knowledge about platform management is high, it is interesting to investigate the different moderating effects.

This research aims to answer the research question by sending out a questionnaire in the Netherlands. This resulted in 206 analysed responses. Accordingly, this research aims to answer the following research question: What are the effects of the consumers’ platform perception and its products on adopting a (potential) platform complementor in the Netherlands?

(11)

10

1.5. Conceptual model

In this study, I test hypotheses of the consumers’ intention to buy a platform complementor presented in the conceptual model of Figure 1. In the following chapter, first, I write about the theoretical background of the model, then, I develop four hypotheses that are indicated in the conceptual model.

Figure 1 Conceptual model of consumers intention to buy a platform complementor

2. Literature review

In this section, I provide a review of the theoretical foundation of this study. I will discuss the various definition of platforms, adoption theories and the concept of perceived ease of use, perceived usefulness and behavioural intention to buy. At last, the network externalities are added and based on argumentation the hypotheses are stated.

2.1. Platforms

Above the definition of platform has been mentioned. Platform, as a business model, is defined as: “Predominant type of business model premised upon bringing different groups together” (Srnicek, 2017). For example, companies as Facebook and Google connect advertisers, businesses, and everyday users in which they apply this business model. This is an addition to the definition given by Gallaugher (2013). This shows not only the importance of complementarity products and services, also the importance of business integration with various partners for the value of platforms. This is beneficial for other firms to contribute to

(12)

11 the platform of a company (Gallaugher, 2013). The integration of these various factors creates value and benefits for the consumer and makes the businesses of platforms as facebook, Amazon and Google rapidly grow. Furthermore, the platforms in the IoT industry are dynamic and innovativeness plays an important role (Choi & Jeon, 2020). Therefore, ICT platforms create new services, which are hardware and software supported innovations this constantly continues in development (Xu, Venkatesh, & Hong, 2010). This leads to multiple generations of platfoms with various consumers, which influences the adoption of a platform.

2.2. Adoption theories

Adoption theories such as Theory of Reasoned Action (TRA) and Technology Acceptance Model (TAM) have gained a lot of results and experience (Ajzen, 1985; Davis, 1989). The conceptual model of this research is derived upon these theories. The TRA explains the relationship between attitude and behaviour of a human action. The TRA points out that the intention to perform a certain behaviour will lead to actual behaviour. The TAM states that two factors are important for acceptation of new innovative technologies: perceived ease of use and perceived usefulness (Davis, 1989). Thus, the higher the perceived usefulness the higher the use.

It is important to first make the distinction between adoption and acceptance as these differ from one and another. Renaud & van Biljon (2008) give the distinction of technology adoption as: “a process, starting with the user becoming aware of the technology, and ending with the user embracing the technology and making full use of it”. Technology acceptance is defined: “an attitude towards a technology, but may not imply a full adoption and use”.

The combination of TRA and TAM are applied in the conceptual model of Gao & Bai (2014).

The conceptual model of Gao & Bai (2014) shows a significant effect between user beliefs and behavioural intention to technology resulting in a single dependent variable: behavioural intention to acceptance of IoT. Behavioural intention is formulated as: “a consumers’ or

(13)

12 employees’ plan to perform specific behaviour” (Davis, 1989). In this light, Kim et al., (2017) defines intention to use as: “a desire to use smart home service”. Next to this, the intention to adopt is defined as: “the degree of an individual’s psychological intent to use a specific system or service” (Davis, 1989).

The variables: perceived ease of use, perceived usefulness and behavioural intention are constructs in many adoption and acceptance studies (Davis, 1989; Guldentops, 2018; Kim et al., 2017). Studies are also applied in the IoT sector and smart home technology sector, with various influential and non-influential factors over the years (Park et al., 2018;

Venkatesh & Davis, 2000). In the following paragraph perceived ease of use and perceived usefulness are explained in detail.

2.3. Perceived ease of use and usefulness

Perceived ease of use is about the consumers’ belief that technology can be used without putting a lot of effort into it (Nikou, 2019; Shin et al., 2018). Next to this, perceived usefulness is about the belief of the user that this system will benefit the user in his or her performance (Venkatesh & Davis, 2000). Consumers belief they will have a positive experience with smart home technologies in daily tasks in household jobs (Fink, Bauwens, Mubin, Kaplan, & Dillenbourg, 2011; Sung, Christensen, & Grinter, 2009). Furthermore, the effects are shown in former studies in which, perceived usefulness (β=0.67, ρ<0.001) and perceived ease of use (β=0.44, ρ<0.001) have a positive effect on behavioural intention to use IoT technologies (Gao & Bai, 2014; Guldentops, 2018; Nikou, 2019). Moreover, studies present that perceived ease of use has a positive relationship towards perceived usefulness.

Meaning, the easier a product or system is to use the more useful it will become (β=0.36, ρ<0.001) (Gao & Bai, 2014; Guldentops, 2018; Nikou, 2019).

This positive significant relationship is influenced by other variables. For example, Venkatesh, Thong, & Xu (2012) indicates three moderating variables: age, gender and

(14)

13 experience. These influence the existing relationship between perceived ease of use and behavioural intention to use. For example, when age of the potential consumer is low the relation between perceived ease of use and behavioural intention of buying a certain

technology becomes stronger. Network externalities are hypothesised to influence the existing relationship between perceived ease of use, perceived usefulness and behavioural intention (Hsu & Lin, 2016; Lin & Bhattacherjee, 2008). Therefore, the next paragraph explains the substantiation for the hypotheses.

2.4. Network externalities

A first example of network externalities research is within computer-mediated communicated (CMC), such as e-mail and other social communication networks (Xiao et al., 2018). In this context, network is not about the physical devices or wires, or wireless systems that connect one and another, but about the digital platform that perform a transaction or communication process. Research towards network externalities expanded, moreover to consumers’ buying behaviour, including the economic effect of network externalities (Lin & Bhattacherjee, 2008;

Lin & Lu, 2011; Yang & Mao, 2014). As mentioned before, network externalities are the effect or value that consumer see, when they use a product or service that is used by more users or a product that is complementary to another used product. The two-sided market and cross-side exchange benefit are two terms to explain as important effect in platform

technology. The reason for this is because they show an effect on the value of a platform (Gallaugher, 2013). This happens when many consumers buy a device from Apple, a certain smart home technology brand. The more consumers use products of this platform, the more attractive this brand becomes. Manufacturers and sellers react on this by making devices possible to connect to this brand and their belonging device, this is called a cross-side exchange benefit in the two-sided market.

(15)

14 Park et al., (2018) shows that benefits that are gained by consumers from a platform are influenced by several network externalities: complementary products and network size. An example of a company in which network externalities are applied is Google. Google is connecting services as e-mail, chat and storage of documents. Moreover, Google Home is a platform in which more products are able to add. For example, Google Nest and Google Chromecast. When more devices are applied this can be seen as a full integrated home as an ecosystem (Hubert, et al., 2019).

In addition to the network externalities, the variables, such as, voluntariness and experiences influence the relationship between perceived ease of use through perceived usefulness to behavioural intention (Venkatesh & Davis, 2000). Meaning, the easier a product or system is to use, the more useful it will become, the higher the intention to buy is, which is strengthend by a high voluntariness.

The last example to give is when a potential consumer wants to buy a smart speaker, it is beneficial when it can integrate with other products owned by the consumer (Park et al., 2018). A smart speaker that is possible to integrate with other devices and applications, increases the intention of a consumer to buy it. The consumer values putting on lights via the smart speaker or asking the smart speaker to read the news. Moreover, the consumer finds it beneficial that more people are using a certain service or product. For example, when a potential consumer sees a smart speaker at a friend’s place, it will increase his/hers intention to buy the smart speaker. Based on the conclusion that platform-related variables are more significant than product-related variables the following hypotheses are described in the next paragraphs.

(16)

15 2.4.1. Network size as a moderator of the direct relationship between perceived ease

of use and the intention of buying

As mentioned before, the relationship between perceived ease of use and intention of buying shows a positive effect. Hence, the easier a system is to use, the higher the intention to buy (Gao & Bai, 2014; Guldentops, 2018; Nikou, 2019). Thus, a product with a high score on perceived ease of use indicates that the potential consumer thinks the product is easy to use resulting in a higher chance adopting products. This relationship can be changed by various variables. The first network externality is perceived network size, when the network size grows, the perceived ease of use will be positively influenced (Park et al., 2018). The number of consumers of a network or platform can be expressed in staying power (Gallaugher, 2013).

A network with a great number of consumers has a stronger staying power. This can influences consumers, because they trust companies with long-term viability. For example, when a consumer is choosing between the operating systems: Windows, Mac OS or Linux.

This is often a choice and investment for over a longer time period. This mechanism of a direct relationship shows the bigger the network, the faster the size of the brand and platform will grow. In the context of the smart home speaker perceived network size has a direct positive relation towards perceived benefits (β=0.245, ρ <0.001) (Park et al., 2018).

Worth-of-mouth marketing is an example of (potential) consumers speak to one and another to evaluate the experiences. Therefore, worth-of-mouth contributes to a larger network size.

This contribution of worth-of-mouth is confirmed by the study of Singh, Kropf, Psychoula &

Hanke (2018) which state that people with many social contacts and high interest in

technology have more acceptance for electronic services at home. Moreover, the study of Lin

& Bhattacharjee (2018) as well as the study of Zhao & Lu (2012) show a positive direct effect of network size on usage (β=0.15, ρ <0.001 and β=0.758, ρ <0.001, respectively).

For example, when you experience the benefits of a smart speaker at a friends’ place, such as ask the speaker to read the news or to put on a music play list, it could increase your intention

(17)

16 to buy the smart speaker. These direct effects are positive effects. In this research, this direct determinant is added as a moderator, because it creates explanatory power (Sun & Zhang, 2006). When the effect of network externalities are positive, it is expected that perceived network size will strengthen the effect between perceived ease of use and behavioural

intention to adopt smart home technology. Meaning that when potential consumers think a lot of people have a platform complementor, they expect that it is easier to learn how to use this product and their intention towards buying this platform complementor will grow stronger.

Based on the above-mentioned arguments the first hypothesis of this research is:

Hypothesis 1 (H1): Perceived network size moderates the relationship between perceived

ease of use and the intention of buying a (potential) platform complementor in such a way that when perceived network size is higher the effect will be stronger.

2.4.2. Perceived complementarity as a moderator of the direct relationship between perceived ease of use and the intention of buying

The direct relationship between perceived ease of use and intention of buying are established above. This paragraph shows the hypothesized positive effect of perceived complementarity towards this relationship. Perceived complementarity can be defined as: “The availability of functions or applications for service” (Hsu & Lin, 2016, p. 520). Park et al., (2018, p. 2122) define this as: “The range of services and supporting applications”. Perceived

complementarity is an indirect externality based on the availability of complementarity services. For example, a smart home technology platform consists of various devices and products allowing, for instance, a smart speaker to communicate with an automatic curtain rails or a robot vacuum cleaner. Another example is that, mulitple complementary

applications can figurate as identification and control access.

Hsu & Lin (2016) state that the life convenience of the user will increase with the number of available complementary products and services. Also, research of Kuebel et al.,

(18)

17 (2015) shows major positive influence of the number of complementarity assets connected to a platform on technology adoption in smart home platforms. Hence, the positive effect or benefit does not only depend on the demand side (number of users) but also the supply side (IoT providers) through the number of complementary products and services (Kuebel et al., 2015).

Furthermore, the indirect network externality can also arise from an indirect participant. For example, a supplier of brand a which adds something to the smart home technology network of brand b. For example, a theromstat from the brand Nest can

communicate with a smart speaker of the brand Samsung, which has a positive influence on both brands. The number of complementary products or services connected to a platform affect the perceived benefit of a product or service (Lin et al., 2011). Moreover, research shows a positive effect of perceived complementarity on the intention of a potential consumer to buy a device (β=0.313, ρ <0.001) (Park et al., 2018).

Finally, a direct, positive and significant effect of perceived complementarity (β=0.33, ρ <0.001) on perceived benefits is presented by Hsu & Lin (2016). Meaning that when

devices and products have a high complementarity to other devices and products the adoption intention will be higher (Lin & Bhattacherjee, 2008). For example, an Apple iPhone has a high complementarity to the Apple computer, because the same operating system and therefore easily synchronise with each other. Hence, the intention of a potential consumer to buy an iPhone will increase.

The second moderator that will be investigated in this research is perceived

complementarity, as the positive relation towards intention to buy is described by the above mentioned various literature sources. When people have the perception that the (potential) platform complementor is working well together with various apps and other devices in their smart home: a high complementarity. This will strengthen the relationship between perceived

(19)

18 ease of use and behavioural intention to buy a (potential) platform complementor. This creates a moderation effect of perceived complementarity towards this direct relation, which results in the following hypothesis.

Hypothesis 2 (H2): Perceived complementarity moderates the relationship between

perceived ease of use and the intention of buying a (potential) platform complementor in such a way that when perceived complementarity is higher the effect will be stronger.

2.4.3. Network size and perceived complementarity as moderators of the indirect relationship between ease of use and the intention to buy

Besides the positive relationship between perceived ease of use and behavioural intention, this research expects also a positive relationship between perceived ease of use and behavioural intention to buy through perceived usefulness (Nikou, 2019; Shin et al., 2018). There is extensive empirical evidence for this indirect relationship (Davis, 1989). The conceptual model of Figure 1 visualises the indirect effect of perceived ease of use through perceived usefulness. This means that the easier a product or system is to use, the more useful it will become and the higher the intention of adopting the technology. Consumers intention to buy a technology is higher when the usefulness or ease of use is high and a combination of a high usefulness and ease of use increases the consumers intention even more. For example, an advertisement of the Ring Video Doorbell that demonstrate how convenient it is to install this product and how easy to use this doorbell is with your phone. This has effect on the intention to buy or use the product. In addition, consumers see that they can answer the doorbell at a distance, which shows the usefulness of the doorbell.

Another example is complementary applications such as electronic tickets and

payments this makes it easier to buy new products or services for users (Hsu & Lin, 2016) and therefore influences the intention of the consumer to buy a new product or service.

(20)

19 This indirect relationship can be influenced by various factors (Venkatesh et al., 2012) for example, as mentioned before age and gender. Network externalities are able to influence this relation as well. The positive effects on perceived ease of use, perceived usefulness and behavioural intention to buy are described above. Because of this reason, a change in the network externalities can strengthen or weaken this relationship. Meaning the higher the perception of network size, the easier the product or system is to use, the higher the usefulness and the higher the behavioural intention to buy. Because of that, the higher the perception of the network size or complementarity of a product, the more positive the influence of

perceived ease of use through perceived usefulness to behavioural intention of buying a (potential) platform complementor. Which results in the following hypotheses.

Hypothesis 3 (H3): Perceived network size positively moderates the indirect relationship of

perceived ease of use towards intention of buying a (potential) platform complementor through perceived usefulness.

Hypothesis 4 (H4): Perceived complementarity positively moderates the indirect relationship of perceived ease of use towards intention of buying a (potential) platform complementor through perceived usefulness.

(21)

20

3. Research design

The proposed conceptual model in Figure 1 is studied in a quantitative nature. Data is gathered through a questionnaire. 231 respondents are questioned on the variables of this research, 206 data points are used in the analysis. This research aims to extend research on potential platform complementors by applying this to the robot vacuum cleaner, which is part of the smart home technology platform, executed in a sample of the Dutch population. As Hubert et al., (2019) advised, this research focuses on a real product that exists in the real market. Next to this, this research focuses on private households. Reason for this is the

evidence that household vacuum cleaners are smaller by size and more cost effective, which is in contrast to the use of vacuum cleaners in the commercial spaces (Fortune Business

Insights, 2020). In 2018 the average of people per household is 2,1 person. Therefore, the theoretical population in 2019 consists of 7.924.691 households (CBS, 2019). As mentioned before, this research gains novelty on the platform-related variables in relation to a specific (potential) complementor of smart home technology. Because of this reason, the respondents are asked about the robot vacuum cleaner.

3.1. Research settings

The literature review showed insights on the various platforms. Platforms are not always about physical wires and devices, but also includes data and communication. This research applies the case of smart home technology and (potential) complementor product.

3.1.1. Smart home technology

With a rise of 29.3% in sales per year from 2018 till 2026 worldwide the expectations are set for smart home technology (Fortune Business Insights, 2020). Smart home technology is defined in various ways. The first definition of “smart home” is described as “an IoT

technology-based system which autonomously produces information and transmits it to other

(22)

21 people and objects, and thereby enhances the quality of residential life” (Kim et al., 2017, p.

1150). Nikou (2019, p. 1) made the following addition on this: “An application of IoT, provides households with e.g. comfort, control and convenience”. In this research the two definitions are combined: providing a household with comfort, control and convenience, with various domestic applications, allowing residents to remotely control, monitor and access their home.

Smart products contain hard- and software such as microchips and sensors these are applied in several products as: automatic lights, self-learning thermostat, smart speakers, remote controlled oven and security of the home via a doorbell with video. Ideally, all these devices are connected to gain the best experience (Hong, Shin & Lee, 2016). It is expected that these platform characteristics trigger the consumers’ intention to buy a smart home product (Park et al., 2018). Therefore, it is no surprise that the consumers’ intention of buying smart home technology is growing. The biggest category is household and comfort with 12%

of the market (Monitor, 2020). The disruptive innovations and benefits of smart home technology seems to become more clear for consumers. Examples of possible benefits are:

increase property value, provide care, enhance leisure, improve quality of life, provide peace of mind and provide comfort (Wilson, Hargreaves, & Hauxwell-Baldwin, 2017). Other benefits mentioned are: assisted living, home security, home entertainment, providing better live standards for elderly, sick and disabled homeowners (Balta-Ozkan, Davidson, Bicket &

Whitmarsh, 2013). The various devices and applications can arise in various sorts, one of these assisting is domestic service robots.

3.1.2. Domestic service robots

Various researchers state that domestic service robots are a promising solution for smart environments, with the following benefits: able to monitor, interact and keep company to users (Nomura & Nakazawa, 2017; Ramoly, Bouzeghoub, & Finance, 2018; Sundmaeker,

(23)

22 Guillemin, Friess, & Woelfflé, 2010). Since recent years it is able to experience this as

autonomous consumer products at home. Current research shows that residents feel the positive effect of practical benefits of a robot and also the emotional component of a domestic service robot (Fink et al., 2011). Next to this, many researchers look into the technical side of service robots, for example the function of making turns, level of communicating and

adapting to learn the map of the house (Forlizzi & DiSalvo, 2006; Prassler, Ritter, Schaeffer,

& Fiorini, 2000). However, it is stated that the robots still need to learn and are adapting.

Service robots can be seen as an interactive device strolling through the house, connecting other smart home devices (Yi Man Li, Ching Yu Li, Kei Mak, & Beiqi Tang, 2016). Another explanation is that a service robot refers to robots which perform useful services to the well- being of humans and equipment (Borja, de la Pinta, Alvares, & Maestre, 2013). Moreover, former research shows that respondents preferred a service robot for less interactive tasks like cleaning (Schiffhauer, Bernotat, Eyssel, Bröhl, & Adriaans, 2018).

3.1.3. Robot vacuum cleaner

The addition of a robot in the domestic environment is not only the physical state or

interactive state of a robot, but also the intelligent network around it (Steijn, Luiijf, & van der Beek, 2016). For example, sensors and artificial intelligence connects multiple devices in the house. This physical robot can be seen as a machine including software making the

possibilities bigger than just one service, for example the service of vacuum cleaning (Steijn et al., 2016). The robot vacuum cleaner has the possibility to vacuum clean the domestic house. The reason for gaining popularity is these light-weighted and cordless vacuum cleaners (Lee, Sung, Sabanovic, & Han, 2012). Moreover, research confirms that the robot vacuum cleaner, is able to provide households with e.g. comfort, control and convenience (Kim et al., 2017; Monitor, 2020; Nikou, 2019; Prassler et al., 2000). The robot vacuum cleaner expects a

(24)

23 rise of 24.4% in sales per year worldwide from 2019 till 2027 (Fortune Business Insights, 2020).

To understand the concept entirely the first definition is of the vacuum cleaner is: “An electrical machine that cleans floors and other surfaces by sucking up dust and dirt”

(Cambridge, 2020). The addition of robot is defined as: “A machine that is controlled by a computer or works automatically” (Cambridge, 2020). Figure 2 shows a picture of the robot vacuum cleaner. The first prototypes of the robot vacuum cleaner have been showed in 1991 (Prassler et al., 2000). The robot vacuum cleaner is an example of a service robot. An

example is that a robot vacuum cleaner can make its own choices regarding making a turn or vacuum more thoroughly on a specific point via this way they can take over labour of a person (Smits & Damen, 2011).

Furthermore, the robot vacuum cleaner has benefits: works without the resident being at home, designed to go under the furniture and clean corners, possible to work on automated times, easy to use, saves time, works on different surfaces. Next to the stated benefits,

challenged need to be solved. Some challenges are: price is higher than standard vacuum cleaner and less strength in vacuuming dirt. Moreover, visual recognition, navigation,

machine learning and a run in a variety of devices need to be strengthened (Gates, 2007). The ethnography study of Forlizzi & DiSalvo (2006) revealed that a vacuum cleaning robot, in contrast to a traditional vacuum cleaner, deployed in the home changed people’s cleaning activities and how they used other tools. The current market of robot vacuum cleaners in the

Figure 2 Picture of a robot vacuum cleaner

(25)

24 Netherlands counts 300,000 households which uses a robot vacuum cleaner and this amount is growing (Monitor, 2020). At last the robot vacuum cleaner is part of the smart home

technology and therefore a single product in a platform environment, important characteristic is that it functions also stand-alone, however this can also be attached to a platform.

3.2. Sample and Procedure

This research is a cross-sectional study on (potential) consumers in the Netherlands.

Respondents are approached via an online questionnaire. Media such as LinkedIn, WhatsApp and Facebook are used for this. The sampling technique is nonprobability sampling, because it does not involve random selection. Convenience sampling is used as the distributing of the questionnaire is among the network of the researcher. Respondents are assured their answers are kept confidential. The questionnaire administration started on October 11^th, 2020 and ended on October 27^th, 2020. To avoid common method variance, several remedies are applied (Chang, Witteloostuijn, & Eden, 2010). The first remedy is about sources, the variables and questions are extracted from different sources. Secondly, different scale measures are used (e.g. slider, matrix and stars). Thirdly, an ex-post is applied by doing a Harman factor analysis and partial correlation procedure.

The questionnaire starts with a textual introduction about the aim of the research, smart home technology and the robot vacuum cleaner. The multi-item scales from previous studies are translated towards Dutch, these are checked by 3 Master’s students on formulation and translation to limit the risk of bias. All variables consists of a 7 point Likert scale. At last the demographic variables are asked to control for. A total of 272 respondents intended to fill in the questionnaire. However, 14 rows of data are recorded but empty and 27 rows of data are not filled in for 100%. This resulted in 231 complete responses. Moreover, respondents

(26)

25 owning a robot vacuum cleaner are excluded from the analyses. 10.8% of the total

respondents owns a robot vacuum cleaner. In total this resulted in 206 respondents.

Prior to further data processing, to assure the anonymity the letters in the postcodes are deleted and changed to provinces. Next to this, descriptive statistics, outlier test, skewness, kurtosis and a normality test are performed. No outliers are found in cleaning the data and missing values are excluded with controlling for only 100% filled questionnaires. Recoding is unnecessary as no reversed questions are used and the questionnaire software is already coded with the correct codes. Descriptive statistics show that of the respondents 79 are male

(38,30%) and 127 are female (61,70%). The range of age is from 17 till 76 (Mage = 38,12, SDage = 14,31). In table 1 the demographic variables can be found. Not all respondents stated their postcodes. All 206 respondents are living in the Netherlands.

Count Table N=%*

Gender* Male 79 38,30%

Female 127 61,70%

Age* < 18 1 0,50%

18-30 92 44,70%

31-40 44 21,40%

41-50 19 9,20%

51-60 29 14,10%

> 61 21 10,20%

Type of living place* Shared house 3 1,50%

Rental house 30 14,60%

Buy house 159 77,20%

Living with parents/caretakers 12 5,80%

Other 2 1%

Province Noord-Brabant 115 59,59%

Gelderland 16 8,29%

Limburg 14 7,25%

Utrecht 14 7,25%

Noord-Holland 13 6,74%

Zuid-Holland 11 5,70%

Overijssel 7 3,63%

Groningen 3 1,55%

Total province 193

*N=206

Table 1 Overview of demographics

(27)

26

3.3. Measures

All measures are derived from English literature. The variables measured in the questionnaire are: perceived ease of use, perceived usefulness, behavioural intention to buy, perceived network size and perceived complementarity. They are adapted in such a way this is focused on the practical instantiation of this research of the robot vacuum cleaner. Perceived ease of use is measured in a scale of Davis (1989) consists of 5 items in a seven-point Likert scale from “strongly disagree”(score 1) to “strongly agree” (score 7) (Cronbach’s Alpha α= .75).

Perceived usefulness is measured in a scale of Davis (1989) consists of 4 items in a Likert scale from “strongly disagree”(score 1) to “strongly agree” (score 7) (Cronbach’s Alpha α=

.90). Behavioural intention to buy is measured in a scale of Venkatesh (2000) consists of 3 items in a Likert scale from “strongly disagree”(score 1) to “strongly agree” (score 7) of (Cronbach’s Alpha α= .93). Perceived network size is measured in a scale of Park et al., (2018) consists of 4 items in a Likert scale from “strongly disagree”(score 1) to “strongly agree” (score 7) of (Cronbach’s Alpha α= .80). Perceived complementarity is measured in a scale of Park et al., (2018) consists of 3 items in a Likert scale from “strongly disagree”

(score 1) to “strongly agree” (score 7) of (Cronbach’s Alpha α= .65).

In table 2 an overview can be found of Cronbach’s Alpha. The reliability coefficients for the measurement items exceed 0.7 except perceived complementarity. Perceived

complementarity is a bit lower than ideal, but the inspection of additional indices, i.e., item- total correlation and Cronbach’s Alpha if the item is deleted, did not point to deletion of a specific item. Therefore, this proceeds with the use of the original scale, but suggested is that future research should focus on further improving the measurement of perceived

complementarity. The questionnaire can be found in appendix 1 in Dutch and the questionnaire items in English can be found in appendix 2.

(28)

27

3.4. Control variables

The first control variables are age and gender. These variables are researched a lot in prior technology acceptance studies in which significant and not significant results are found (Ezer, Fisk, & Rogers, 2009; Fink et al., 2011; Guldentops, 2018; Nikou, 2019; Sung et al., 2009).

Literature is not conclusive about the differences between male and female, neither younger nor older persons. Therefore, this research controls for these two variables. Moreover, the third control variable is the type of living place of a potential consumer, which is advised to include by Hall, Backonja, Painter, Cakmak, Sung, & Demiris (2019). Operationalising this term means there is controlled for (potential) consumers living in an apartment, rental house or buy home.

At last, the postcode is asked, which is transformed into provinces after the data is gathered. The first reason for asking the postcode is to control the sample for spread throughout the Netherlands. The second reason is controlling for the place of living.

These variables are analysed and results are forthcoming in the next chapter.

(29)

28

4. Results

The analysis of the data is processed via IBM SPSS Statistics. First, the data is prepared for analysing. Second, data is processed in normality, validity and reliability checks. Third, correlation and factor analysis will be made. Finally, the hypotheses are tested with the Macro Process of Andrew F. Hayes.

4.1. Normality and factor analysis

The data is tested for normality and validity. Appendix 3 includes the complete plots and output for these tests. First, the p-value for both the Kolmogorov-Smirnov test and the Shapiro-Wilk test is smaller than the 0.05 probability level. This suggest that the data is not normally distributed. However, the histograms and Q-Q plots show a relatively normal distribution for the focal variables: behavioural intention to buy, perceived usefulness and perceived complementarity. To assess the risk of the common method bias, the Harman’s test is used. Namely, factor analysis, all items are set to load on a single factor. The results

indicate that such single factor solution explain 32.60% of the variance, which is below the recommended 50% threshold. Concluding, the results of these tests does not confirm a considerable threat in terms of common method bias.

Second, in table 2 the output of skewness and kurtosis can be found. Four of the variables show negatively skew and one of the variables shows a positively skewed, showing not normal distribution. The sample must be representative to avoid sample selection bias (Trochim, Donnelly, & Arora, 2016). The sample used in this research is drawn from the researchers environment which could threat the credibility. This is shown in the demographic variables, almost 60% of the sample lives in Noord-Brabant. Because of that reason external validity should be treated with care. However, “with reasonably large samples, skewness will not make a substantive difference in the analysis” (Tabachnick & Fidell, 2001 p. 74).

“Kurtosis can result in an underestimate of the variance, but the risk is also reduced with a

(30)

29 large sample” (Tabachnick & Fidell, 2001 p.74). In this research more than 200 participants therefore the risk is reduced and skewness would not make a substantive difference in the analysis.

Construct

Number of

items Cronbach's Alpha

Scale

improvement Skewness Kurtosis

4. Intention to buy 3 (.93) no .207 -.96

5. Perceived Usefulness 4 (.90) no -.082 -.88

6. Perceived Ease of Use 5 (.75) no -.915 1.035

7. Perceived Complementarity 3 (.65) no -.111 -.338

8. Perceived Network Size 4 (.80) no 1.149 2.050

Table 2 Summary of reliability results

4.2. Correlation analysis

Table 3 shows an overview of the descriptive statistics, correlations and scale reliabilities.

Perceived network size significantly positively correlates with behavioural intention to buy (r

= 0.25, p < 0.01); positively not significantly correlates with perceived usefulness (r = 0.12, p

> 0.05), and positively significantly with perceived ease of use (r = 0.18, p < 0.05). Perceived complementarity significantly positively correlates with behavioural intention to buy (r = 0.39, p < 0.01); positively correlates with perceived ease of use (r = 0.33, p < 0.01) and perceived usefulness (r = 0.39, p < 0.01). These results indicate that perceived

complementarity is associated with all variables of the conceptual model: perceived ease of use, perceived usefulness and behavioural intention to buy and perceived network size. This does not imply causality. Nevertheless, perceived network size is associated with none of the variables of the conceptual model. No significant correlation appears between perceived network size and perceived usefulness. Moreover, perceived network size is significant with a level of p < 0.05 towards perceived ease of use, showing less stronger results than other significant relationships. Only the control variable age is showing a negative significant

(31)

30 relationship, meaning the higher the age the lower the score on the other variable. Other control variables do not show any significant relationships.

Variables M SD 1. 2. 3. 4. 5. 6. 7. 8.

1. Gender 1.62 0.49 /

2. Age 38.12 14.31 -.10 /

3. Type of living place 2.90 0.54 -.05 .01 /

4. Intention To Buy 3.40 1.58 -.01 -.32** .04 (.93)

5. Perceived Usefulness 4.20 1.40 -.09 -.29** .07 .72** (.90)

6. Perceived Ease of Use 5.28 1.17 -.02 -.35** .06 .52** .50** (.75) 7. Perceived Complementarity 4.12 1.20 .00 -.25** -.10 .39** .39** .33** (.65) 8. Perceived Network Size 2.21 0.97 .01 -.24** -.01 .25** .12 .18* .23** (.80)

**. Correlation is significant at the 0.01 level (2-tailed).

*. Correlation is significant at the 0.05 level (2-tailed).

Note. The scale reliabilities (alphas) are displayed along the main diagonal

Table 3 Means, Standard Deviations, Correlations

Prior to testing hypotheses and interpreting the results, assessment for

multicollinearity takes place. The outcomes can be found in table 4. All factors of tolerance are greater than 0.2 meaning there is no collinearity problem. VIF is less than 5.0 meaning there is limited evidence of multicollinearity problem when other indicators of the problem are unserved. Some variables have an eigenvalue close to 0 meaning that the variables are highly intercorrelated, therefore showing a small problem towards multicollinearity.

Moreover, the condition index is higher than 15, indicating the presence of multicollinearity.

Variable Collinearity Statistics

T Sig Tolerance VIF

1. Gender 0,759 0,449 0,974 1,027

2. Age -0,762 0,447 0,809 1,236

3. Type of living place 0,052 0,959 0,972 1,028

5. Perceived Usefulness 10,594 0,000 0,676 1,48

6. Perceived Ease of Use 3,069 0,002 0,689 1,452

7. Perceived Complementarity 1,379 0,170 0,778 1,286

8. Perceived Network Size 2,644 0,009 0,909 1,100

N = 206 * p<0,05, ** p<0,01

a: Dependent variable is intention to buy Table 4 Summary of test for multicollinearity

(32)

31

4.3. Hypotheses testing

Moderator analysis is performed to test the hypothesized relationship of perceived network size and perceived complementarity on the relationship between perceived ease of use and perceived usefulness on behavioural intention to buy a robot vacuum cleaner. When

performing moderation analysis, the analysis controlled for the role of different demographic variables: age, gender and type of living place. All hypotheses are tested using PROCESS macro for SPSS with 10.000 bootstrap samples under Model 1 and 7. Outputs of these analyses can be found in appendix 4.

4.4. Moderation effects

The first hypothesis is tested by using model 1, the results can be found in table 5. The model as a whole is significant, (R²=0.311, F (6.199) = 15.033, p<0.000). The regression coefficient for XM is C3=-0.03, this means the effect of perceived ease of use on the intention to buy a robot vacuum cleaner does not depend on the perceived network size. Perceived ease of use (β=0.52, p<0.001) significantly predicts intention to buy, but the effect of perceived network size on this relation is not significant. The variables together explain 31,12% of the variance (R²=0.311).

The second hypothesis is tested by using model 1, the results can be found in table 6.

The model as a whole is significant, (R²=0.342, F (6.199) = 17.201, p<0,000). The regression coefficient for XM is C3=-0.02 this means the effect of perceived ease of use on the intention to buy a robot vacuum cleaner does not depend on the perceived complementarity. The model explains 34,15% of the variance (R²=0.341). Perceived ease of use has the strongest effect in the model. Perceived network size and perceived complementarity do not affect the relation between perceived ease of use and behavioural intention to buy. There is no evidence found to support hypotheses 1 and 2 and therefore the hypotheses will be rejected.

(33)

32 Moderation effect of perceived network size

Coefficient SE T p

Intercept i1 -0.06 1.28 -0.05 >0.005

Perceived ease of use (X) C¹ 0.65 0.19 3.42 <0.005

Perceived network size (M) C² 0.37 0.48 0.78 >0.005

perceived ease of use * perceived network size (XM) C3 -0.03 0.08 -0.31 >0.005

R² = 0,3119 p<0,000

F(6,199) = 15,03

Table 5 Output hypothesis 1

Moderation effect of perceived complementarity

Coefficient SE T p

Intercept i1 -0.79 1.41 -0.56 >0.005

Perceived ease of use (X) C1 0.60 0.24 2.52 <0.005

Perceived complementarity (M) C² 0.41 0.32 1.27 >0.005

Perceived ease of use * perceived complementarity (XM) C³ -0.02 0.06 -0.32 >0.005 R² = 0,3415 p<0,001

F(6,199) = 17,20

The third hypothesis is tested by using model 7, the results can be found in table 7.

This is a moderated mediation analysis. The model as a whole on perceived usefulness is significant, (R²=0.271, F (6.199) = 12.39, p<0.000). Next to this, the model on intention to buy is significant (R²=0.56, F (5.200) = 51.91, p<0.000). These explain 27,19% and 56% of the variance. The results further indicate no effect of perceived ease of use on intention to buy a robot vacuum cleaner on perceived network size, as no evidence is found between XW in the model of Y (A3=-0.02, p>0.0005).

Perceived usefulness

(M) Intention to buy (Y)

Coeff. SE p Coeff. SE p

Perceived ease of use (X) A¹ 0.56 0.17 <0.005 C¹ 0.26 0.07 <0.005

Perceived usefulness (M) B¹ 0.69 0.07 <0.005

Perceived network size (W) A² 0.10 0.44 >0.005 Perceived ease of use * Perceived

network size (XW) A³ -0.02 0.08 >0.005

Constant i¹ 1.87 1.18 >0.005 i² -0.65 0.68 >0.005

R² = 0.2719

F (6,199) = 12,39, p<0.000

R² = 0.56 F (5,200) = 51,91,

p<0.000

(34)

33 The fourth hypothesis is tested by using model 7, the results can be found in table 8.

This is a moderated mediation analysis. The model as a whole on perceived usefulness is significant, (R²=0.33, F (6.199) = 15.98, p<0.000). Next to this, the model on intention to buy is significant (R²=0.56, F (5.200) = 51.91, p<0.000). These explain 33% and 56% of the variance. The results further indicate no effect of perceived ease of use on intention to buy a robot vacuum cleaner on perceived complementarity, as no evidence is found between XW in the model of Y (A3=0.04, p>0.005).

Perceived usefulness

(M) Intention to buy (Y)

Coeff. SE p Coeff. SE p

Perceived ease of use (X) A¹ 0.31 0.17 >0.005 C¹ 0.26 0.07 <0.005

Perceived usefulness (M) B¹ 0.69 0.07 <0.005

Perceived complementarity (W) A² 0.09 0.29 >0.005 Perceived ease of use * Perceived

complementarity (XW) A³ 0.04 0.05 >0.005

Constant i¹ 1.63 1.27 >0.005 i² -0.65 0.68 >0.005

R² = 0.33

F (6,199) = 15,98, p<0.000

R² = 0.56 F (5,200) = 51,91,

p<0.000

(35)

34 Thus, the effect of perceived ease of use and perceived usefulness is the same for all levels of perceived network size and perceived complementarity. This means no significant moderation effect happens on the relation towards behavioural intention of buying a robot vacuum cleaner. In table 9 an overview of the hypotheses is visualised. However, it is predicted that a high perceived complementarity or high perceived network size strengthens the direct and mediating relationship.

Hypotheses Outcome

H1. Perceived network size moderates the relationship between perceived ease of use and the intention of buying a (potential) platform complementor in such a way that when perceived network size is higher the effect will be stronger.

Rejected

H2. Perceived complementarity moderates the relationship between perceived ease of use and the intention of buying a (potential) platform complementor in such a way that when perceived complementarity is higher the effect will be stronger.

Rejected

H3. Perceived network size positively moderates the indirect relationship of perceived ease of use towards intention of buying a (potential) platform complementor through perceived usefulness.

Rejected

H4. Perceived complementarity positively moderates the indirect relationship of perceived ease of use towards intention of buying a (potential) platform complementor through perceived usefulness.

Rejected

Table 9 Summary of hypotheses

Master thesis Are Network Externalities the Key for Consumers to Adopt Platform Technology?

Master thesis

Are Network Externalities the Key for Consumers to Adopt Platform Technology?

A Quantitative Study on the Moderating Role of Network Externalities on the Potential Consumers’ Buying Intentions of Platform Technology

Statement of Originality

Table of contents

Abstract

1. Introduction

2. Literature review

3. Research design

4. Results