Adoption of social networking sites in different industries

(1)

Adoption of social networking sites

in different industries

An empirical analysis of Facebook

Lauren C.M. Bembom | S3186318

University of Groningen

Faculty of Economics & Business

MSc BA Small Business & Entrepreneurship

Supervisor: dr. M.J. Brand

Co-assessor: dr. A.J. Frederiks

Keywords Social Networking Sites Adoption, Facebook Adoption, Industries, TOE Framework JEL Classifications C19, D22, J21, L00, L25, O33

(2)

2

Abstract

(3)

3

1. Introduction

1.1 Introduction

The World Wide Web is evolving. The first generation of internet sites started as information place for firms to broadcast their information to people and provided limited user interactions (Aghaej, Nematbakhsh, Farsani, 2012). With the rise of sites such as Facebook, the web is becoming increasingly interactive. On this Web 2.0, content is no longer created by individuals, but instead continuously modified by all users (Kaplan & Haenlein, 2010).

Social media build on the technological foundations of Web 2.0 and bring a new element into the promotion mix of firms (Kaplan & Haenlein, 2010). Traditionally, promotion enables firms to talk to their customers, while social media enable customers to talk directly to one another (Mangold & Faulds, 2009, p. 359). Therefore, customers tend to have more trust in online recommendations about a product, service or brand than in traditional advertisements (Van Looy, 2015, p. 2). Social media “influence various aspects of consumer behaviour, including awareness, information acquisition, opinions, attitudes, purchase behaviour and evaluation” (Mangold & Faulds, 2009, p. 358). This consumer behaviour could lead to potential benefits for the business, for example “increased customer focus and understanding, increased level of customer service, and decreased time-to-market” (Jussila, Kärkkäinen & Leino, 2012).

Social networking sites (SNS) are a type of social media and are defined as “applications that enable users to connect by creating personal information profiles, inviting friends and colleagues to have access to those profiles, and sending e-mails and instant messages to each other” (Kaplan & Haenlein, 2010, p. 63). Currently, hundreds of millions of users have adopted SNS (Martins, Gonçalves, Pereira, Oliveira & Cota, 2014). Adoption is “the decision to make full use of an innovation” (Rogers, 1995, p. 21). Only few studies of SNS adoption exist. The majority is about adoption at individual level, but adoption at firm level remains neglected in the literature (Martins et al., 2016). At firm level, studies try to explain information technologies (IT) such as electronic data interchange (Kuan & Chau, 2001), websites (Beatty, Shim & Jones, 2001; Oliveira & Martins, 2008) and e-commerce (Liu, 2008; Martins & Oliveira, 2009).

There are new challenges for SNS compared to previous IT, such as lack of information control, interactivity with consumers and low cost and high efficiency marketing (Mangold & Faulds, 2009; Saldanha & Krishnan, 2012; Kaplan & Haenlein, 2010). Consequently, SNS can be considered as a new class of IT (Martins, Gonçalves, Oliveira, Cota & Branco, 2016). Factors influencing adoption by the firm may be different from that of previous IT. These factors may still be influential, but relative importance to the adoption process may differ (Martins et al., 2016). Because the domain of SNS is relatively new, there is a lack of theory testing and theory building associated with SNS (Mallouli, Hachicha & Chaabouni, 2017, p. 136). Because literature did not provide a clear perspective on SNS at firm level, Martins et al. (2016) developed a conceptual framework with the Delphi method and thorough analysis of existing literature. Factors were identified that influence SNS adoption in three contexts: top management, organization and environment.

(6)

6

Chamber of Commerce (Van der Meer, Smetsers & De Jong, 2017) and an Australian study (Sensis, 2016). For example in the Netherlands, 40% of the firms in the ‘Agriculture, forestry and fishing’ industry have adopted SNS against 76% of the firms in the ‘Accommodation and food service activities’ industry (Van der Meer et al., 2017). These studies suggest that firms focused on the endconsumer (B2C) use SNS more often than firms in the business to business (B2B) industries. Yet, these studies are not scientific and replicability is low. The adoption of social media in different industries is not well understood in the literature (Jussila et al., 2014, p. 2). Also, no scientific research exists about what factors may explain industry differences in SNS adoption, in the next paragraph this knowledge gap will be explained.

Commonly used data collection methods in SNS adoption studies are surveys and interviews (Oliveira & Martins, 2011). Many of these SNS studies use self-reported usage which can lead to common method bias (Tiago & Veríssimo, 2014; Rauniar, Rawski & Johnson, 2014; Lee, Kozar & Larsen, 2003). Also, response behaviour may be influenced as people are aware of the fact that they are being studied. Besides that, studies do exist where interpretations do not play a role: programs analyse “social media big data” (Tufekci, 2014, p. 505). This data mining is now widely seen as the approach to study online communication, because “it has the aura of being easy, comprehensive, and objective” (Varis, 2014, p. 16). However, collecting and getting access to social network data, computational power and algorithmic accuracy are problems that arise (Neunendorf, 2016; Varis, 2014; Van Looy, 2016). Because programs miss contextual understanding, they can misinterpret data (Boellstorf et al., 2012; Chui et al., 2012; Varis, 2014). Therefore, there is need for an objective data collection method in the field of SNS adoption. In the next paragraph this will be explained.

1.2 Problem statement

It seems that SNS adoption certain industries is almost twice as much as in other industries. Since research

explaining differences between SNS adoption among industries is lacking in the literature, this study is important to solve the knowledge gap.

If industries indeed differ in SNS adoption, it is relevant to take industry into account when studying

(7)

7

1.3 Research questions

Based on identified knowledge gap and the goals of the study, two research questions are formulated: RQ 1: ‘What is the actual SNS adoption across industries?’

RQ 2: ‘How can differences in SNS adoption between industries be explained?’

The two main goals of this study are: a) to get insight into actual SNS adoption across industries. This study focuses on Facebook, because it is the most widely used social networking site (Alalwan et al., 2017; Kaplan & Haenlein, 2010; Rauniar et al., 2014); b) to identify factors that explain differences in SNS adoption between industries. Furthermore, goals of this study are c) to develop a concrete guide to help firms use SNS effectively d) to collect data objectively as one of the first in the SNS adoption research field. Although this study focuses on Facebook adoption, outcomes may be generalised to SNS.

1.4 Practical and academic relevance

The theoretical relevance of this study is that it provides insight in actual SNS adoption across industries and on how differences in SNS adoption between industries can be explained. This is relevant because if research is carried out in which industry is not included, results will not be valid. This study is theoretically relevant as objective data collection methods will be seen as worthy data collection methods in the SNS adoption literature.

The practical relevance of this study is that with the results, firms can benchmark themselves with the average firm in their industry. As already described, SNS can be valuable for the firm, however, because SNS are open for every firm to use, it can be seen as a common resource. According the Resource Based View, valuable but common firm resources can be exploited to create competitive parity in an industry (Barney, 1989a). Although no firm obtains competitive advantage, “firms do increase their probability of economic survival” (Barney, 1991, p. 107). Some firms do not know how to use social media, which can lead to problematic SNS use that may negatively impact a firm’s business (Patel, 2016; Patterson, 2015). Insights from this study can give guidance to firms on how to use SNS in their specific industry

1.5 Structure of thesis

(8)

8

2. Literature Review

This chapter elaborates literature that tries to explain technology adoption and introduces industries and their characteristics. These industry characteristics may explain SNS adoption differences and can be linked to the DOI model (Rogers, 1995) and the TOE framework (De Pietro, Wiarda & Fleischer, 1990). Furthermore, this chapter formulates hypotheses and develops a conceptual model.

2.1 Technology adoption

In the literature various studies try to explain technology adoption. These studies can be divided into two broad categories: studies that attempt to explain technology adoption at the individual level and at the firm level. The majority of SNS adoption studies are at the individual level, whereas firm level adoption remains scant (Grawe, Daugherty & Richey, 2010; Martins et al., 2016). At the individual level, The Technology Adoption Model (Davis, 1980), as well as the Unified Theory of Acceptance and Use of Technology model (Venkatesh, Morris, Davis and Davis, 2003) and Theory of Planned Behaviour (Azjen, 1985) are the most widely used models for investigating factors affecting technology adoption (Saldanha & Krishnan, 2012; Martins & Oliveira, 2011). The basis of these models comes from social psychology and determinants of a user’s intention to adopt a technology are ‘perceived usefulness’ and ‘perceived ease of use’.

The adoption process in an organization is more complex, because it often involves a number of individuals, who can be supporters or opponents of new technology and each of whom plays a role in the innovation process (Oliveira & Martins, 2011). At the firm level, the Diffusion of Innovation Theory (DOI) of Rogers (1989) and the Technology-Organization-Environment Framework (TOE) of De Pietro et al. (1990) are the most widely used models. To a lesser extent, the Institutional Theory (Scott, 1987) and the Resource-Based Theory (Wernerfelt, 1984) are used.

In DOI theory, the decision to adopt or reject the innovation depends on the characteristics of the decision-making unit, as well as on the perceived characteristics of the innovation (Rogers, 1989). The TOE framework (DePietro et al, 1990) comprises the same components as the DOI model, as it also includes characteristics of the decision-making unit (named Organizational context), and perceived characteristics of the innovation (named Technological context) (Gorla, 2017). The TOE framework additionally included the Environmental context, see Table 1. The TOE framework is often used to describe the context in which adoption takes place and incorporates factors that influence adoption by organizations based on the Technological, Organizational and Environmental context (Sharif, Troshani & Davidson, 2014). The DOI theory and TOE framework are used to examine adoption of wide range of technologies such as electronic data interchange (Kuan & Chau, 2001), websites (Oliveira & Martins, 2008) and e-commerce (Martins & Oliveira, 2009).

Table 1

Components in innovation adoption models

DOI theory (Rogers, 1989) TOE Framework (De Pietro et al., 1990)

Perceived characteristics of the innovation Technological context

Characteristics of the decision-making unit Organizational context

(9)

9

2.2 SNS: a new kind of technology

Social media are “a group of internet-based applications that build on the ideological and technological foundation of Web 2.0” that are “freely accessible and where users meet each other for the purpose of creating, modifying, discussing and exchanging matters together” (Kaplan & Haenlein, 2010, p. 61; Van der Meer et al.,

p. 5). Various types of social media need to be distinguished. Social media can be classifiedinto blogs (e.g.,

Frankwatching.com), virtual social worlds (e.g. Second Life), collaborative projects (e.g. Wikipedia), content communities (e.g. YouTube), virtual game worlds (e.g. World of Warcraft) and SNS (e.g. Facebook) (Kaplan & Haenlein, 2010). SNS are defined as “applications that enable users to connect by creating personal information profiles, inviting friends and colleagues to have access to those profiles, and sending e-mails and instant messages to each other” (Kaplan & Haenlein, 2010, p. 63). Some studies add that users cannot only connect to friends and colleagues, but also with others to whom they do not have a real life affinity (Pai & Arnott, 2013; Haythornthwaite, 2005).

SNS are consistent with ‘traditional’ IT such as e-mail, websites and e-commerce, as companies can use SNS to talk to their customers (Mangold & Faulds, 2009). But SNS have also unique properties that make them “uniquely powerful enablers of value creation”, because it enables social interactions “with speed, scale and economics of the internet” (Chui et al., 2012, p. 3). In Table 2 properties of ‘traditional’ IT and SNS are shown. Because of these distinct properties of SNS compared to previous IT, SNS can be considered as a new class of IT (Martins et al., 2016).

Table 2

Properties of ‘traditional’ technologies and SNS technology

Properties Traditional technologies SNS technology Author

Communication (Sometimes)

interactive Interactive Saldanha & Krishnan, 2012 Mangold & Faulds, 2009;

Goal of use Focus mainly on

consumption Focus mainly on establishing and

maintaining social relationships

Mata & Quesada, 2014; Parveen, 2012; Michealidou

et al., 2011; Kaplan & Haenlein, 2012

Word of mouth To several people To hundreds of people Mangold & Faulds, 2009

Degree of information

control High Low Kaplan & Haenlein, 2010; Mangold & Faulds, 2009

Cost High Low Kaplan & Haenlein, 2010;

Leeflang et al., 2014; Alalwan et al., 2017

Level of efficiency Low High Kaplan & Haenlein, 2010;

Leeflang et al., 2014; Alalwan et al., 2017

2.3 SNS adoption

(10)

10

interchangeably (Martins et al., 2016; Perrigot et al., 2012; Michealidou et al., 2011). Perrigot et al. (2012) defines SNS adoption as ‘setting up a presence on Facebook’. Most studies only look if the firm has a presence on social media (Perrigot et al., 2012). Some studies also look at the level of social media usage in terms of number of posts and number of social media platforms (Van der Meer et al., 2017).

Martins et al. (2016) developed a conceptual framework for SNS adoption at firm level. This included a Delphi study with 25 experts in the IT/IS area and a thorough and systematic analysis of earlier research on IT adoption. Experts scored a list of factors influencing SNS adoption on relative importance. With the ten most important factors a research model was designed and hypotheses were tested through a survey with a response of 247 firms. Four factors were identified with most impact on SNS adoption: “Top Management Support”, “Alignment of SNS plan with Business plan”, “Use of SNS for Competitive Advantages” and “Competitive pressures on firms to use SNS”.

2.4 SNS adoption across industries

Defining the industry in which competition takes place is important for good industry analysis (Porter, 2008, p. 14). It is increasingly difficult to define precisely where an industry begins and ends (Hamel & Prahalad, 1996). Defining the industry too broadly “obscures differences among products, customers, or geographic regions that are important to competition, strategic positioning, and profitability”, whereas defining the industry too narrowly “overlooks commonalities and linkages across related products or geographic markets that are crucial to competitive advantage” (Porter, 2008, p. 14).

Porter (2008, p. 14) identified two boundaries of the industry: the scope of the product or service, and the geographic scope. If the industry structure of two firms is the same or very similar (same buyers, suppliers, barriers to entry), the firms can be treated as being part of the same industry (Porter, 2008). The International Standard Industrial Classification (ISIC) is the international reference in which firms can be classified on productive activities. The main goal of the ISIC is to provide a set of activity categories that can be utilized for the collection and reporting of statistics according to such activities.

In SNS studies, if any distinction in industry is made, it is typically between ‘service’ and ‘goods’ Michealidou, et al., 2011; Perrigot et al., 2012). These studies argue that the primary goal to adopt SNS is to develop brand equity, which is more important for tangible goods than for services. Mixed results are found: Michealidou et al. (2011) found no significant differences between firms operating in services and goods industries with respect to their SNS adoption; Perrigot et al. (2012) found significant differences in SNS

adoption between franchisors in the service sector and the retail sector. Maybe because of this broad

distinction, little empirical evidence is found for differences between industries in SNS adoption. Yet from several recent non-scientific studies it seems that in percentages, firms in certain industries adopt social media more than in other industries, see Table 3. However, it is not yet studied if this difference is significant. Therefore, the following hypothesis is formulated:

(11)

11

2.5 Industry characteristics

Scholars highlight the importance of industry as critical contextual variable with regards to IT impact and suggest that different sectors would be impacted in different ways (Stoel & Muhanna, 2009; Chui et al., 2012). If social media is valuable for firms in a certain industry, depends on “fundamental characteristics of the industry” (Chui et al., 2012, p. 10; Stoel & Muhanna, 2009). Van der Meer suggest that at least following three firm characteristics influence social media adoption, which may also reflect characteristics of the industry: 1. Target market. Some entrepreneurs argue that social media does not match the customers and products

of their company (Van der Meer et al., 2017). The lack of perceived relevance of SNS in particular industry sectors is the main reason why firms do not adopt SNS (Michealidou et al., 2011). Firms targeted on the end-consumer seem to use social media more often than firms in the business to business industries (Van der Meer et al., 2017).

2. Firm size. An indicator for available resources can be firm size (Thong, 1999). In the literature is known that limited number of employees can form barriers to adoption of innovations (Larsen & Lewis, 2007). CBS data (Appendix 1) and the study of Van der Meer et al. (2017) indicate that, generally, the higher the number of employees, the more social media is used. For each industry, the average number of employees differ (Appendix 2). Because of the fact that the average number of employees differ per industry and the higher the number of employees, the higher SNS adoption, it is plausible that the number of employees explain part of the industry difference in SNS adoption.

3. Industry pressure to use SNS. Data indicate that the more social media is used by the competition, the more intensive social media is used by the firm (Van der Meer et al., 2017). The competitive environment pressures companies to adopt and use SNS (Martins et al., 2016). Each industry has its own competition. Therefore, the competitive pressure may be a characteristic of the industry explaining SNS adoption differences.

Table 3

Adoption of social media by % firms in industry

Van der Meer et al., 2017 Sensis, 2016

76 %: Accommodation and food service activities 76 %: Education

73 %: Culture, sports and recreation

71 %: Renting, buying and selling of real estate

67 %: Wholesale and retail trade, repair of motor vehicles and motorcycles 65 %: Renting and leasing of tangible goods

62%: Other service activities

60 %: Information and communication 58 %: Human health and social work activities 53 %: Manufacturing

52 %: Consultancy, research and other specialised business activities 49 %: Financial institutions

44 %: Construction

42 %: Transportation and storage 40 %: Agriculture, forestry and fishing

Adoption of social media in the Netherlands. SBI 2008 Classification. Corresponds to European NACE Rev 2 & ISIC Rev 4 (world level).

77 %: Cultural, recreational and personal service

61 %: Retail 60 %: Hospitality 60 %: Communication,

property and business services 35 %: Manufacturing

32 %: Building and construction

(12)

12

2.6 Explaining SNS adoption differences

The next paragraphs formulate hypotheses for the three industry characteristics. Because the TOE framework (DePietro et al.,1990) is widely used to explain technology adoption, the industry characteristics will be grouped in one of the three contexts (Technology, Organizational and Environmental context).

2.6.1 Factors in the Technological context

The first industry characteristic ‘target market’ can be linked to the Technological context of the TOE framework. The technological context concerns the characteristics of a technology that influences its adoption within organizations (DePietro et al., 1990). According to a literature review of Hoti (2015) factors as ‘relative advantage’, ‘complexity’ and ‘compatibility’ are most often identified in this context as having impact on technology adoption. ‘Compatibility’ is the “degree to which an innovation is perceived as consistent with existing values, past experiences and adopter needs” (Hoti, 2015, p. 7).

The business value of social media primarily lies in customer engagement (Stockdale, Ahmed & Scheepers, 2012). It is argued that several properties of SNS technology align better with business values of firms targeted on individuals (B2C) than of firms targeted on businesses (B2B). The American Marketing Association defines B2B firms as “business which markets their products to other businesses” in contrast to B2C firms which “sell products directly to individual consumers” (Lacka & Chong, 2015). The properties of SNS

technology may influence the perceived usefulness of SNS adoption by firms, in other words, in order for SNS

to be adopted, it should be “efficient in getting tasks done” (Rauniar et al., 2014).

Firstly, SNS support establishment of virtual relations between individuals, between organizations, and between individuals and organizations, but individuals are the “first and most important beneficiaries of social technologies” (Martins et al., 2016; Chui et al., 2012, p. 11). This is because on SNS, users need to create a personal profile to make use of the platform: all profiles on SNS are managed by individuals, profiles of firms are managed by (few) employees of the firm with their personal (private) profiles (Kaplan & Haenlein, 2010; Martins et al., 2016). Therefore, firms that target on individuals (B2C, business to consumer market), have a higher chance to reach its target market on SNS than firms that target on other firms (B2B, business to business market). Secondly, it is difficult to build trust in an online environment as SNS and opportunistic behavior is harder to avoid (Bridges, Goldsmith & Hofacker, 2005). Co-operation is more direct and intense in the B2B market than in the B2C market (Jussila et al., 2014). Generally, in the B2B market there is often intense co-operation between firms and it is important to build and maintain trust. Furthermore, the person controlling the SNS business page may not be the person which deals with business partners. Following transaction cost theory, a B2B firm would prefer an ‘offline’ relationship with their business partner. Thirdly, B2B has more specific demands and fewer buyers compared to the B2C market (Gorla, 2017; Jussila et al., 2014; Lacka & Chong, 2015). Buyers in the B2B market have more emphasis on physical performance and personal selling than in consumer products, where psychological attributes and advertising are critical for success (Urban & Hauser, 1993).

(13)

13

market and “the main reason for firms' reluctance to use SNS for particular industry sectors” (Michealidou et al., 2011, p. 18). Studies support that firms in the B2C market adopt social media more often than firms in the

B2B market (Lacka & Chong, 2015; Van der Meer et al., 2017). Following literature and the reasoning above, it

seems that SNS is less relevant for businesses targeting on businesses, a bit more relevant for firms targeting on both consumers and business and most relevant for firms targeting on consumers. Therefore, the following hypothesis is formulated:

H2. Targeting on the consumer market positively influences SNS adoption.

2.6.2 Factors in the Organizational context

The second industry characteristic ‘firm size can be linked to the Organizational context of the TOE framework (DePietro et al., 1990). The organizational context concerns descriptive measures of the firm, such as size, scope, structure and resource availability" (DePietro et al., 1990; Oliveira & Martins, 2011, p. 112). Organizational readiness is often mentioned as significant factor in the organizational context that influence technology adoption (Alshamaila, Papagiannidis & Li, 2013; Evangelistam Esposita, Lauro & Raffa, 2010, Ghobakhloo, Arias-Aranda & Benitez-Amado, 2011). In these studies, organizational readiness encompass organizational size and their corresponding available resources that influence technology adoption: if the firm has the right resources available to easily use the technology (perceived ease of use) (Larsen & Lewis, 2007). These resources can be used to aid and enhance the adoption of technologies (Hoti, 2015). Larger firms usually possess human and financial resources, whereas in SMEs these resources are limited (White, Afolayon & Plant, 2014).

For IT, due to resource poverty, small firms face more difficulties in the adoption of innovations than large firms and therefore adopt IT at lesser extent than large firms (Hoti, 205; Burgess et al., 2015; Rahayu & Day, 2016). In order to adopt SNS, limited financial resources are needed (Kaplan & Haenlein, 2012;

Michealidou et al., 2011; Leeflang, Verhoef, Dahlströ & Freundt , 2014). However, limited scale of human

resources can still form barriers to adoption of innovations (Larsen & Lewis, 2007). Human resources are the ‘pool of human capital under the firm’s control in a direct employment relationship’, also known as the employees of the firm (Wright, McMahan, McWilliams, 1994, p. 6). According to the CBS, 61% of the business with 10-20 employees have adopted SNS, whereas 91% of the businesses with 500 or more employees have adopted SNS (Appendix 1). Studies argue that even though social media can be adopted by all sizes of firms, the “continual usage will be more prominent in large firms due to resources available to them” (Parveen, 2012, p. 7). Prior literature identified the number of employees as one of the most important determinants of IT adoption (Love et al., 2005; Premkumar, 2003; Thong and Yap, 1995). Also a recent study found that franchise systems that are present on Facebook, have significantly more employees than franchise systems that were not present on Facebook (Perrigot et al., 2012). The larger the firm, the more SNS is used (Appendix 1). As industries differ in firm size (Appendix 2), the firm size may explain industry differences in SNS adoption. In order to test this, the following hypothesis is formulated:

(14)

14

2.6.3 Factors in the Environmental context

The third industry characteristic ‘industry pressure’ can be linked to the Environmental context of the TOE framework of DePietro et al. (1990). The environmental context concerns the industry, competitors, and the regulatory environment of the firm (DePietro et al., 1990). ‘Industry pressure’ is a factor often mentioned in the TOE literature that has influence on technology adoption in the environmental context (e.g. Ghobakhloo et. al., 2011; Parker & Castleman 2009; Scupola, 2009; Haug et. al. 2011; Doom, MIlis, Poelmans, Bloemen, 2010). The economic environment in which firms work constantly changes and this causes pressures on firms to constantly adapt (Martins et al., 2016). “As firms start to adopt and use SNS, a competitive pressure starts to arise forcing firms that still have not adopted it, to exert a great deal of effort to start using” (Martins et al., 2016, p. 18).

Streams in the literature have different views on competition in industries. Models like Porter (1985) assume that firms within an industry are identical in terms of strategically relevant resources they pursue, and if resource heterogeneity develops in an industry, it is only temporarily because the resources are highly mobile. However, Resource Based View argues that firms within an industry may be heterogeneous with respect to their strategic resources they control and that this heterogeneity can be long lasting as resources may not be perfectly mobile across firms (Barney, 1991). A firm can achieve sustained competitive advantage if a firm resource possesses following 4 attributes: a) it is valuable, in a sense that it exploit opportunities/neutralizes threats in a firms environment, b) it is rare among a firm’s current and potential competition c) it is perfectly imitable and d) there are no strategically equivalent substitutes for this resource that are valuable but neither rare or perfectly imitable.

A resource is valuable when “they enable a firm to conceive of or implement strategies that improve its efficiency and effectiveness” (Barney, 1991). SNS allow firms to have contact with consumers at “relatively low cost and higher levels of efficiency than can be achieved with more traditional communication tools” (Kaplan & Haenlein, 2010, p. 67), and therefore SNS can be valuable resource for a firm. Martins et al. (2016) found a significant relationship between use SNS for gaining a competitive advantage and SNS adoption. The advantages addressed are lower costs, electronic links with suppliers or customers, influence users decision and leverage firm capabilities. However, if the valuable firm resources, the SNS platform, is possessed by large numbers of competing firms, it cannot be sources of either a competitive advantage of sustained competitive advantage” (Barney, 1991, p. 106). Still, these valuable and common resources can help ensure a firm’s economic survival when they are exploited to create competitive parity in an industry (Barney, 1991).

In the literature is a significant relationship found between the existence of competitive pressure and SNS adoption (Martins et al., 2016). The more social media are used by competitors, the more social media are used by the firm (Van der Meer et al., 2017). To test this in practice, the following hypothesis is formulated:

(15)

15

2.7 Conceptual model

The conceptual model is shown in Figure 1. H1 checks if industries actually differ in SNS adoption. If this is the case the variables target market (H2), firm size (H3) and industry pressure to use SNS (H4) are added. These variables may explain part of the industry effect. If H1 is still significant, then there are other relevant industry characteristics, this is ‘remaining industry effect’ and may be interesting for future research.

Figure 1. Conceptual model Technological context

Target market

Organizational context Firm size

Environmental Context Industry pressure to use SNS

(16)

16

3. Methodology

This chapter describes how the empirical part of the study is carried out. This chapter elaborates the research plan, data collection method, measures and discusses the quality of the study.

3.1 Research approach

In Chapter 1, the main goals of the study are formulated, namely to get insight into actual SNS adoption across industries and to identify factors that explain differences in SNS adoption between industries. This research follows the quantitative theory testing approach. In explanatory research sciences, the empirical cycle is usually followed. On the basis on literature research, we develop a conceptual model, translate this model into empirically testable hypotheses and then test these hypotheses (Vennix, 2011; Van Aken et al., 2012). After this, results will be analysed, interpreted and a conclusion will be drawn.

3.2 Data collection method

As already described in Chapter 1, data collection methods used by the majority of studies in SNS literature such as surveys and interviews, have an important disadvantage: subjectivity. Further, although big data analysis makes computational counts, it may miss the context. In order to strive for an objective description and explanation of reality, this study tries to collect ‘hard’ data. All data needed for this study are already available in existing sources and therefore the disturbance in the situation being investigated is minimalized. Existing data sources can be grouped into edited (secondary) data and primary data (Vennix, 2012, p. 192). This study collects both types of data. We collect data from the Orbis database and observe content on the firm’s website and Facebook. Orbis is a database that contains financial and business information from 200 million companies worldwide. Orbis strives to get their database of firms as complete as possible, and among others data is based on information from the Chamber of Commerce. We also collect additional variables which are not essential to answer the main question, but help to replicate the research and may be interesting for further analysis. Data will be collected on firm level and later will be combined to arrive at statements at industry level. The data collection procedure is shown in Figure 2.

Figure 2

(17)

17

3.3 Sample

This study wants to make statements about Dutch firms in different industries. Therefore, the empirical population of the research consist of all registered companies at the Chamber of Commerce (KVK) in the Netherlands. The KVK distinguishes 21 main sectors (CBS ‘Standaard Bedrijfsindeling 2008, v. 2018’). This study will focus on 15 industries, which consist of 1% or more of the total number of firms in the Netherlands. This focus is similarly to the study of Van der Meer et al. (2017), however this study uses a subjective data and our study use relatively objective data. Results of both studies can be compared on discrepancies.

It is expected to have some chains (franchisor/ headquarter of a chain) or chain units (franchisee or one of the locations of a chain) in the sample. The chain units may adopt SNS themselves, but sometimes only the chain has SNS. In this case, the chain and chain units post on the chains’ SNS. Chain units that adopt SNS themselves may take advantage of ready-made content from the chain. Because of this strategic cooperation, chains and their units are expected to adopt SNS more than other firms. The influence of chains will be taken into account in the analysis.

For this study, a stratified sample will be performed. First subpopulations, based on industry, are made and then a sample is taken from each subpopulation. As this study wants to make statements about firms in different industries, a non-proportional sample will be taken within each stratum, to adequately represent the smaller industries that are relatively underrepresented in the sample (Vennix, 2011, p. 83). The sample is a-select, with each element of the population having an equal chance to end up in the sample. Afterwards these samples are combined into a (stratified) sample from the entire empirical population.

To test a hypothesis in which the influence of an independent variable on a dependent variable is examined, the probability must be normally distributed. According to the Central Limit Theorem, the sample distribution is (almost) normal when the sample size is large enough (n ≥ 30) (McClave & Sincich, 2011). To ensure that the sample is normally distributed, the sample size of the Central Limit Theorem is taken into account. This means 15 industries x 30 firms per industry will be analysed: a total of 450 firms.

The selection procedure for the sample in Orbis is listed in Table 4. This selection will be made for each industry. Then, a random sample of 100 companies will be selected by Orbis. If no information about the company is found (website link broken and no information found in Google search on company name), or a firm is classified into the wrong industry, this will be noted and the subsequent company will be selected from the random sample until for each industry, results from 30 firms are collected. The subsectors ‘Financial holdings’, ‘stichting administratiekantoren’ and investment institutions are different legal ‘entities’ that only represent part of a firm in order to spread risk. These entities bias results and will therefore be filtered out.

Table 4

Sample selection procedure in Orbis

1. Status: Active companies and companies with Unknown situation 2. Geographic: World region/Country/Region in Country: Netherlands

3. Number of employees: All companies with a known value, Last available year, exclusion of Public authorities/States/Governments

4. Contact information: All companies with a website address

5. NACE Rev 2 Industry Main Section [selection of one of the 15 industries].

(18)

18

3.4 Measures

In order to make the transition from a conceptual model to empirical observation, variables need to be operationalized (Vennix, 2011, p. 143). This section will explain how variables from the conceptual model will be measured and what additional data will be collected. For a quick overview, see Appendix 3A.

Demographic variables

• Company name. The name of the company as stated in official documents.

• Industry of the firm. The ISIC (International Standard Industrial Classification) is the world standard from the United Nations. The NACE of the European Union is based on the first two digits of the ISIC. The last two digits provide European detailing. The Dutch standard industrial classification (SBI) base the national activity classification SBI on the four digits of NACE. At the industry level, the European NACE Rev 2 and the Dutch SBI 2008 correspond to ISIC Rev 4 (world level) (CBS ‘Relaties tussen (inter)nationale standaardclassificaties’). Because this classification is worldwide used to define industries, this study also adopts this classification. The 15 included industries are: A) Agriculture, forestry and fishing; C) Manufacturing; F) Construction; G) Wholesale and retail trade, repair of motor vehicles and motorcycles; H) Transportation and storage; I) Accommodation and food service activities; J) Information and communication; K) Financial institutions; L) Renting, buying, selling and real estate; M) Consultancy, research and other specialised business activities; N) Renting and leasing of tangible goods and other business support services; P) Education; Q) Human health and social work activities; R) Culture, sports and recreation; S) Other service activities.

The excluded industries are: B) Mining and quarrying; D) Electricity, gas, steam and air conditioning supply; E) Water supply sewerage, waste management and remediation activities; O) Public administration, public services and compulsory social security; T) Activities of household as employers; undifferentiated goods- and service-producing activities of households for own use; U) Extraterritorial organisations and bodies. These industries are very small: only a total of 4.030 firms from 6 industries (0.24% of all registered Dutch firms) are excluded from the empirical population. To improve readability, industry names will be partly abbreviated.

• URL Website. The company name as stated in official documents does often not correspond to the name of the company known by the public (trade name). The link of the website of the company often encompass the trade name, therefore this is a requisite in the sample selection in Orbis.

• URL Facebook. The link to the social networking site Facebook-page of the company (if available). • Chain/chain company. Chains/chain units: it is expected these firms post more than ‘regular’ firms. • Additional interesting variables: Subsector: to check if company is in right industry; Province, Region, City:

(19)

19

Variables in the conceptual model

• SNS adoption. The presence of the firm on Facebook (has an official Facebook page) (Perrigot et al., 2010). • Level of SNS adoption. Many studies do not measure the level of SNS adoption, but this may be important as it gives more nuance. Several authors measure the level of adoption in terms of frequency of social media used by the user (never, rarely, occasionally, often, frequently) (Rauniar et al., 2014), or number of posts in the categorisation no SNS & barely, daily, weekly or monthly (Van der Meer et al., 2017). The level of SNS adoption will be measured in usage of Facebook of the firm (number of posts on official Facebook page at www.facebook.com). Usage is counted from January 1, 2017 to January 1, 2018: a whole year because of possible seasonal differences/influences. This also includes firms with no SNS (measured as no usage).

• Target market. Different target markets can be distinguished, namely firms targeted on businesses (B2B, firm sells to supply chain partner organization (Gorla, 2017), business and consumers (B2B/B2C, firm sells to supply chain partners and individuals) (Anwar, 2017) and individuals (B2C, firms sells to individual/consumer (Gorla, 2017). Each firm is assigned to one of the three codes. This will be done by observing firms’ content placed on the website and (if available) Facebook. Because of interpretation, the researcher may bias results. Therefore, a codebook is used as it gives more reliable results. A few studies in SNS literature observed SNS content (Perrigot et al., 2012, Mergel, 2013, Parveen, 2012), however, no instructions or coding schemes can be found. Consequently, a codebook needs to be developed (Appendix 3B). A codebook includes detailed instructions what and how to code (Perrigot et al., 2012). • Firm size. To measure firm size, the number of employees in a company is commonly used as measurement

in management literature (Drechsler, Wenzel, Natter, Leeflang 2013; Meiseberg, Ehrmann, 2013; Lee, Kozel & Larsen, 2003). The number of employees are ‘the pool of human capital under the firm’s control in a direct employment relationship’ (Wright et al., 1994, p. 6) and can be found in Orbis.

• Industry pressure to use SNS. As one item to measure competitive pressure, Martins et al. (2016) ask the respondents if competitors use SNS. In our study, for each industry the percentage of SNS adoption and the average level of SNS adoption are computed to z-scores. The industry pressure is the average of both z-scores.

3.5 Data analysis method

(20)

20

3.6 Quality of the study

This section discusses the quality in terms of reliability and internal and external validity of the study. The reliability is the overall consistency over a measure and is high reliable if it produces similar results under consistent conditions. This study is highly reliable, data collection steps are clearly defined and all data can be found online in the database Orbis and on Facebook. If the same steps are followed, the same results can be reproduced.

The internal validity is the degree to which conclusions about causal relationships can be made based on the measures used. Factual data will be collected such as number of employees, company name, URL website/Facebook, ‘the company has Facebook yes/no’ and counting the number of posts. Defining the target market and the content of the posts may encompass some interpretation of the researcher, but is minimalized by using a codebook. Furthermore, concepts are derived from existing research, such as SNS adoption and industry pressure. Therefore, the internal validity is high.

(21)

21

4. Results

This chapter describes how the research was conducted and shows the results. 4.1 Sample

In total, 104 firms of the original sample did not fulfill the requirements and are replaced. For the industry ‘Financial institutions’ the sample of 100 firms was too small and a new sample of 200 firms was randomly selected. Also, in the ‘Financial institutions’ industry sample three locations of Rabobank were selected. This is unrepresentative to reality and will bias results. Therefore two locations of the Rabobank were replaced with new cases. All dataset decisions can be found in Appendix 10. The aim was to collect data of 450 firms and this goal is achieved. The sample is representative for the empirical population, see Appendix 4.

4.2. Results

This section presents the results for each hypothesis. Because dependent variable SNS adoption is measured in two ways, namely SNS adoption as presence/absence and as the level of SNS adoption (in number of posts), each hypothesis is tested twice. The first step in the analysis is to clean the data by filtering out outliers and checking assumptions as linearity, homoscedasticy, independence and normality (Field, 2013). If the assumptions are met, descriptives will be elaborated and a Chi-square test, logistic regression, one-way ANOVA and hierarchical regression analysis will be performed. The hypotheses are then tested and the results described.

4.2.1. Outliers

Outliers can bias a parameter estimate such as the mean and influence the error associated with that estimate (Field, 2013). Using z-scores, the main variables target market, number of employees, industry pressure, SNS adoption and the level of SNS adoption are checked for outliers (z-scores higher than ± 3.29). 11 outliers are found for firm size (with z-scores of 3.4 to 9.2), from which 4 chain units. For level of SNS adoption, 12 firms are found with z-scores from 3.3 to 8.5), from which 7 chain units and 1 chain. The cases that contain outliers are deleted as a whole, as they bias results. A total of 427 firms remain in the dataset.

4.2.2. Checking assumptions

Before statistical analyses can be done, it is important to check if no assumptions are violated (Pallant, 2001).

Table 5

Descriptive statistics

Mean Median Skewness Kurtosis

Firm size 4.73 1.00 3.524 13.985

Industry pressure to use SNS 0.0 -0.206 0.908 0.290

(22)

22

Firm size, industry pressure to use SNS and level of SNS adoption are continuous variables see Table 5. For these continuous variables the mean, the median, the skewness and kurtosis are checked (Pallant, 2001). Positive kurtosis values indicate that the distribution is clustered in the centre with long thin tails (Pallant, 2001, p. 54). With reasonably large samples (200+ cases) this makes no substantial difference in the analysis (Tahbacnick & Fidell, 1996). The parameter estimates of the population will have a normal distribution as the samples are approximately ‘big enough’ (widely accepted is a value of 30).

The variables target market, SNS adoption and industry are categorical and should be examined

whether groups are equal or unequal. Groups are considered equal when the number in the largest group to the number in the smallest group is <1.5 (Hair, Black, Babin & Anderson, 2010, p. 459). Groups in the target market are equal (157 (B2C)/ 128 (B2B)=1.23). Groups in SNS adoption are unequal (257 firms did adopt/ 170 firms did not adopt = 1.51). Industry groups can be considered as equal (30/26=1.2).

4.2.3 Descriptives

Firms, chains and chain units

Table 6 shows SNS adoption for firms, chain units and chains. In total, there are 427 companies in the dataset, from which 32 chain units and 7 chains. The high adoption rate of in total 87,5% for chain units (v.s. 58.2% of firms) may be explained by the strategic cooperation with the chain: chain units can adopt themselved and/or can rely on the chain’s SNS and therefore do not have to adopt themselves. Chain (units) will be included in the sample as this are also companies that present themselves on SNS. The influence of chains on SNS adoption will be discussed in ‘additional insights’

SNS Adoption

Table 7 shows SNS adoption among industries. In terms of percentage, it can be said that industries differ in SNS adoption. Table 7 also shows a comparison with the study of Van der Meer et al. (2017), this will be further elaborated in the discussion.

Table 6

SNS adoption firms, chain units and chains

Firms Chain units Chains

% N % N % N

No adoption 41.8 162 12.5 4 57.1 4

Own SNS adoption 58.2 226 21.9 7 - -

Chain SNS adoption - - 65.6 21 42.9 3

(23)

23 Level of SNS adoption

Table 8 shows the level of SNS adoption. Industries that adopt SNS the most in percentages, do not necessarily have the highest level of SNS adoption. For example, the industry Education is in the top five highest SNS adopters, whereas the industry has one of the lowest levels of SNS adoption. When including chains and chain units, in some industries in which chains and units are present, the usage almost doubles. Chain (units) will be included as they are also present on Facebook and compete against other firms. The exact influence of chain (units) on the level of SNS adoption will be elaborated in the ‘additional insights’.

Table 7

SNS adoption

Industry This study (presence) SNS adoption Social media adoption Van der Meer et al,

2017 (>11 posts)

Difference

Rank % adoption Rank % adoption % difference

Accommodation and food 1 93.3 1 76 + 17.3

Other service activities 2 75.9 7 62 + 13.9

Wholesale and retail trade 3 75.0 5 67 + 8

Education 4 67.9 2 76 - 8.1

Culture, sports and recreation 5 66.7 3 73 - 6.3

Human health 6 65.5 9 58 + 7.5

Real estate 7 61.5 4 71 - 9.5

Business support services 8 57.7 6 65 - 7.3

Agriculture, forestry and fishing 9 53.3 14 40 + 13.3

Transportation and storage 10 51.9 15 42 + 9.9

Construction 11 50.0 13 44 + 6

Information and communication 12 50.0 8 60 - 10

Manufacturing 13 50.0 10 53 - 3

Financial institutions 14 46.4 12 49 - 2.6

Consultancy 15 35.7 11 52 - 16

Table 8

Level of adoption in average usage

Industry chains/ units Usage

included Usage chains/units excluded # of firms chain # of units # of

chains sample Total

Accommodation and food 51.53 53.44 27 2 1 30

Human health 34.93 31.74 27 1 1 29

Culture, sports and recreation 33.87 33.87 30 0 0 30

Wholesale and retail trade 33.79 30.83 24 4 0 28

Real estate 30.42 24.82 22 2 2 26

Other service activities 25.21 17.60 25 4 0 29

Financial institutions 23.89 14.23 22 6 0 28

Business support services 20.46 12.17 23 2 1 26

Manufacturing 19.86 16.61 23 4 1 28

Consultancy 16.11 15.67 27 1 0 28

Information and communication 11.50 11.90 29 0 1 30

Transportation and storage 10.48 8.96 24 3 0 27

Education 10.29 9.67 27 1 0 28

Agriculture, forestry and fishing 8.80 8.80 30 0 0 30

Construction 4.37 4.68 28 2 0 30

(24)

24

4.2.4 Hypotheses

Hypothesis 1 is first tested for SNS adoption and the level of SNS adoption. Then Hypothesis 2, 3 and 4 are tested for both. After each hypothesis, some additional insights are described in which industries are grouped in higher and lower (level of) adoption, services and goods, and main target market.

Hypothesis 1 [SNS adoption]

To investigate the relationship between the two categorical variables ‘industry’ and ‘SNS adoption’, the Pearson’s Chi-square test of independence is used. The sample meets the assumptions of independence and expected frequencies (minimum expected count is 10.35; none of the cells have an expected count less than 5). The Chi-square test compares per industry the frequencies observed for ‘no SNS adoption’ and ‘SNS adoption’ to the expected frequencies in those categories by chance (Field, 2013, p. 722). Table 9 presents the percentage of total ‘no SNS adoption’ and ‘SNS adoption’ for each industry and the corresponding standardized residual expressed in a z-value. The standardized residual, (z-value) is the error beteen expected and observed frequency. Z-values outside ±1.96 deviate significantly and contribute significantly to the overall Chi-square statistic. The expected counts are (almost, because of slight sample size differences) the same for each industry, assuming that there are no differences in SNS adoption between industries. Chi-square does not compares the counts itself, but the proportions.

For df=14, the critical value for the Chi-square distribution is 29.14 for p=.001. There is a significant

relationship between industries and whether or not SNS is adopted

c

2 _{(14)=35.329, p=.001). This association is}

mainly driven by firms in the ‘Accommodation and food’ industry (they adopt SNS significantly more than expected, z=2.3, 60.2 % of the sample do adopt SNS vs. 93.3% within this industry) and firms in the ‘Consultancy’ industry (they adopt significantly less than expected, z=2.1, 39.8% of the firms in the sample do not adopt SNS, within this industry 64.3%). Therefore, Hypothesis 1 ‘SNS adoption differs across industries’ is

Table 9

Results Pearson’s Chi-square test

Industry No SNS Adoption SNS Adoption

Expected Observed

z-value Expected Observed z-value

Accommodation and food 11.9 2 2.9** 18.1 28 2.3*

Other service activities 11.5 7 - 1.3 17.5 22 1.1

Wholesale and retail trade 11.1 7 - 1.2 16.9 21 1.0

Education 11.1 9 - .6 16.9 19 .5

Culture, sports and recreation 11.9 10 - .6 18.1 20 .5

Human health 11.5 10 - .5 17.5 19 .4

Real estate 10.4 10 - .1 15.6 16 .1

Business support services 10.4 11 .2 15.6 15 -.2

Agriculture, forestry and fishing 11.9 14 .6 18.1 16 - .5

Transportation and storage 10.7 13 .7 16.3 14 - .6

Construction 11.9 15 .9 18.1 15 - .7

Information and communication 11.1 15 .9 18.1 15 - .7

Manufacturing 11.1 14 .9 16.9 14 - .7

Financial institutions 11.1 15 1.2 16.9 13 - .9

Consultancy 11.9 18 2.1* 18.1 10 - 1.7

(25)

25

accepted. For the remaining industries SNS adoption is as expected. Cramer’s V is preferred to measure the strength of the relationship as variables have more than two categories (Field, 2013, p. 725). Cramér’s V is significant (p=,001) and is .29 out of a possible maximum of 1, which represents a weak association between the type of industry and SNS adoption.

Additional insights

See Appendix 6. Firms in the ‘Accommodation and food’ industry are 25 times more likely to adopt SNS than firms in the ‘Consultancy’ industry. For other industries, the odds ratios can be found in Appendix 6A. When

grouping industries in high and low adoption industries, Chi-square is significant,

c

2 _{(1)=23.802, p=.000,}

meaning there is a significant relationship between industry and SNS adoption (Table B). When grouping the

industries in services/goods industries, no significant relationship is found with SNS adoption (Table C),

c

2

(1)=2.185, p=.139. If grouped in B2B, B2B/B2C and B2C industries, B2B industries adopt significantly less than expected and B2C firms adopt SNS significantly more than expected (Table D). Furthermore, a chi-square test

comparing chain (units) with other firms on SNS adoption found

c

2 _{(2)=6.672, p=.010. Chain (units) adopt SNS}

more than expected (Table E). The odds ratio is 2.8, meaning chain (units) are 2.8 times more likely to adopt SNS than other firms, however this difference is not significant.

Hypothesis 1 [Level of SNS adoption]

Next the hypothesis is tested for the level of SNS adoption, which is the average usage of the firm of SNS. As the dependent variable is continuous and the independent variable is categorical, a one-way ANOVA is performed. ANOVA is used to analyze differences among group means (industries) on the level of SNS adoption. Because the assumption of homogeneity is violated (no equal variances), the Welch statistic is used. The results are shown in Table 10, indicating a significant difference between industries in level of SNS adoption. Therefore, Hypothesis 1 ‘SNS adoption differs across industries’ can be accepted. A more detailed version is in Appendix 7.

In order to check which industries differ significantly from each other on the level of SNS adoption, 105 significance tests (K (k-1) / 2) are performed. Because of heterogeneity, the Games-Howell Post Hoc test is performed (Appendix 7B). Firms in the industries ‘Accommodation and food’ and ‘Construction’ differ significantly on the level of SNS adoption. The firms in the ‘Accommodation and food’ industry differ almost

significantly from ‘Agriculture’, ‘Transportation’ and ‘Education’. R2_{=.08, a less biased estimator of the variance}

explained is omega squared= ω2_{= .06. The industry differences on SNS adoption not are not very substantive,}

as .5 is the threshold for a large effect (Field, 2013, p. 472).

Table 10

ANOVA. Level of SNS adoption

Dependent variable Independent variable W Sig. ω2 _Effect

size

(26)

26 Additional insights

See Appendix 7. With industry grouped on low and high level of adoption, industries differ significantly from each other (Table C). When goods and service industries are compared at the level of SNS adoption, no significant differences are found (Table D). When grouping industries in B2B, B2B/B2C and B2C industries, the B2C industries do significantly differ from B2B industries and from B2B/B2C industries on the level of SNS adoption (Table E). Furthermore, with an independent t-test is checked if chain (units) do differ on the level of SNS adoption, which is found to be significant (Table F).

Hypothesis 2, 3 and 4 [SNS adoption]

Binary logistic regression is used when the dependent variable is dichotomous. Logistic regression predicts the probability of SNS occurring given known values of several predictor variables. In order to test if the variables target market, number of employees and industry pressure significantly explain differences in SNS adoption among industries, binary logistic regression is performed. Dummies are made for the categorical variable industry. The assumptions of logistic regression (linearity and independence of errors, overdispersion

parameter ϕ =.88, smaller than 1) are met. Also, there is no multicollinearity (VIF tolerance not bigger than 5,

tolerance is from .810 to .936 (not lower than .01).Results of the regression are summarized in Table 11.

Model 1 predicts 64,9% of the outcomes correctly, c2_{(3)=55.615, p=.000. Adding industry to the model}

is not significant, Model 2: c2_{(13)=8.559, p=.805, indicating that the variables firm size, target market and}

industry pressure to use SNS significantly predict differences between industries in SNS adoption. These three variables all have a significant positive relationship with SNS adoption (respectively Exp(B)=1.87, Exp(B)=1.04

and Exp(B)=1.59). Therefore, H2 ‘Targeting on the consumer market positively influences SNS adoption’,H3 ‘The

firm size positively influences SNS adoption’ and H4 ‘Industry pressure to use SNS positively influences SNS adoption’ can be accepted. The model is significant, the overall fit of the model is -2LL=518.48, but a well-fitting model will have a small value for -2LL (Hair, Anderson, Tatham & Black, 1998). The effect size of Model 1 is measured

in Hosmer and Lemeshow’s measure: !2_#= .097, where 0 indicates that predictors are not very useful at

(27)

See Appendix 8. A logistic regression is performed to analyse the influence of chains on SNS adoption. Model

1 predicts 64,9% of the outcomes correctly, c2_{(3)=55.615, p=.000. Adding chain to the model is almost}

significant, Model 2: c2_{(1)=3.577, p=.059 (Table A). Also a logistic regression is performed for industries}

grouped on high and low adoption (Table B). Model 1 predicts 64,9% of the outcomes correctly, c2_(3)=55.615,

p=.000. Adding low/high adoption industries to the model is not significant, Model 2: c2_{(1)=55.753, p=.710}

(Table B).

Hypothesis 2, 3, 4 – Level of SNS adoption

In order to test if target market, firm size and industry pressure to use SNS significantly explain the level of SNS adoption, multiple (hierarchical) regression is used. Dummies are used for the categorical variable industry. The assumptions of linearity and homoscedasticy are met (predict slightly better for low number of posts than high number of posts). There is independence of predictor variables (no multicollinearity) as the average VIF is not substantially higher than 1 and tolerance is above .2.

The linear model of predictors of the level of SNS adoption are shown in Table 12. Model 1 explains a significant part of the variance in the dependent variable level of SNS adoption F(3, 423)=14.277, p=.000. Industry does not make a significant addition to the model F=(13, 410)=3.106, p=.877. The ANOVA indicates that the model as a whole is significant, F(3,423)=14.277, p=.000. Therefore, Hypothesis 2 ‘Targeting on the

consumer market positively influences SNS adoption’ and Hypothesis 4 ‘Industry pressure to use SNS positively

influences SNS adoption’ are accepted. Firm size does not significantly predict the level of SNS adoption and

therefore Hypothesis 3 ‘The firm size positively influences SNS adoption’ is rejected. The model has an explained

variance of .108 which indicates a weak relationship between independent variables and the dependent variable.

Table 11

Logistic regression. SNS adoption

b (95% CI) SE p _$%

& Sig. F Change

Model 1 0.097 .000 Constant -0.985 .33 .003 Target market .626 (1.412, 2.48) .14 .000 Firm size .037 (1.01, 1.02) .02 .012 Industry pressure .464 (1.22, 2.07) .13 .001 Model 2 0.111 .805 Constant 5.176 .831 Target market 0.742 (1.53, 2.78) 0.16 .000 Firm size 0.038 (1.01, 1.07) 0.02 .013 Industry pressure -8.974 (0.00, 3.03) 35.69 .801 Industry (N=15) Summarized in: Adoption rank 1-5 Adoption rank 6-10 Adoption rank 11-15 3.709 -7.756 -13.746 0.07 29.13 51.90 .787 .675 790

Model 1: R2_{=.10 (Hosmer & Lemeshow) .12 (Cox & Snell) .17 (Nagelkerke), Model 2: R}2_{=.11 (Hosmer &}

(28)

See Appendix 9. Chain is added in the model as predictor (Table A). In Model 1, target market, firm size and industry pressure are included and explain significant part of the SNS adoption. F(3, 423)=14.277, p=.000. In Model 2, chain is added and explains a significant part of industry differences in SNS adoption F(1, 422)=14.550, p=.000. In Model 3, industry is added but does not make a significant addition to the model F(13, 409)=3.817, p=.876.

Instead of the fifteen industries, the regression is also performed for the industries grouped in high and low level adoption (Appendix 9B). Here, In Model 1, target market, firm size and industry pressure are included and explain significant part of the SNS adoption. F(3, 423)=14.277, p=.000. In Model 2, higher and lower level adoption industries are added, but are no significant addition to the model F(4, 422)=10.916, p=.000.

Table 12

Linear model of predictors of the level of SNS adoption.

b (95% CI) SE β p R2 _{Sig. F Change}

Model 1 .092 .000 Constant 6.699 (-5.64, 19.04) 6.28 - .287 Target market 6.574 (1.23, 11.88) 2.70 .13 .015 Firm size 0.444 (-0.03, 0.91) 0.24 .09 .064 Industry pressure 10.160 (5.59, 14.73) 2.32 .22 .000 Model 2 .108 .877 Constant -1.322 (-18.74, 16.10) 8.86 - .881 Target market 8.329 (2.48, 14.17) 2.97 .16 .005 Firm size 0.469 (-0.02, 0.96) 0.25 .09 .060 Industry pressure 12.039 (5.80, 18.28) 3.18 .26 .000 Industry (N=15) Summarized in:

Level of adoption rank 1 - 5

Level of adoption rank 6 - 10 4.542 7.667 10.11 9.51 .03 .04 .645 .395

Level of adoption rank 11-15 2.129 10.25 .01 .584

(29)

29

5. Discussion and Conclusion

This chapter gives an answer to the central questions. Furthermore it elaborates theoretical and practical implications and the quality of the research. Finally, this chapter gives suggestions for future research. 5.1 Discussion

5.1.1 Actual SNS adoption across industries

It seems that 93.3% of the firms in the ‘Accommodation and food’ industry have adopted SNS, whereas only 35.7% of the firms in the ‘Consultancy’ industry have adopted SNS. Using the Chi-square, these industries do differ significantly from the other industries. When industries are classified in a low-adoption and a high-adoption group, they differ significantly in SNS high-adoption.

Industries also significantly differ in the level of SNS adoption. Firms in the industry ‘Accommodation and food’ use SNS on average 51.5 times a year, whereas firms in the ‘Construction’ industry use SNS on average 4.4 times a year, which is a significant difference. The industries ‘Agriculture’, ‘Transportation’ and ‘Education’ almost differ significantly from the ‘Accommodation and food’ industry. When industries are grouped in a high and a low level of adoption group, they differ significantly in the level of adoption. Remarkable is that industries that adopt SNS the most, do not necessarily use SNS the most. For example the industry ‘Education’ has a relatively high adoption (rank 4 out of 15) whereas it has a relatively low level of adoption (rank 13 out of 15). It is likely that a higher amount of non-profit firms is present in the ‘Education’ industry.

Our study found that service industries adopt and use SNS as much as goods industries, which is in

conformity with Michealidou et al. (2011). B2B, B2B/B2C firms differ significantly in SNS adoption, we already

argued that the relevancy for customer targeted firms is higher. Previous SNS studies implicitly suggest that

B2C firms adopt social media more often than B2B firms, but to our knowledge it is never compared (Lacka & Chong, 2015; Van der Meer et al., 2017). B2B/B2C and B2B firms do not differ in usage, but differ significantly from B2C industries: B2C firms use SNS around 17 extra times a year. Chains and chain units are 2.8 times more likely to adopt SNS than other firms, which is slightly insignificant. Chain (units) use SNS average 29 extra times per year than other firms and this difference is found to be significant. This may be explained by the strategic cooperation between the chain and chain units.

5.1.2 Explanation of differences in SNS adoption between industries

Three industry characteristics are defined in order to try to explain (level of) SNS adoption differences among industries. Target market and industry pressure significantly explain (the level of) SNS adoption. This means the more the firm targets on consumers, the higher (the level of) SNS adoption. In particular industries in which targeting on B2C is most present, the highest adoption rates are found. A possible explanation is the fit with the technology with consumers: previous literature also argued that SNS is more suitable for B2C firms than

for B2B firms (Martins et al., 2016; Chui et al., 2012). Targeting plays an important role in the adoption and use

Adoption of social networking sites in different industries