• No results found

Online shopping & Social media

N/A
N/A
Protected

Academic year: 2021

Share "Online shopping & Social media "

Copied!
60
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

From attitudinal SM E-loyalty to behavioral SM E-loyalty. Passive and active social media engagement as a mediating bridge in the social media context.

MAX TUK S3700623

University of Groningen Faculty of Economics and Business

MSc Marketing Management June 13, 2020

Eewal 86A 8911GV Leeuwarden m.tuk.1@student.rug.nl

First supervisor: Dr. J.I.M. de Groot Second supervisor: Dr. J. Hans Berger

(2)

Abstract

Customer loyalty is a well-known and popular topic in marketing literature, which can be viewed as a multidimension construct that consists of an attitudinal and a behavioral

component. It is assumed that behavioral intentions, originating from attitudinal loyalty could positively affect actual behaviors and hence increase behavioral loyalty. Within this study, the interrelatedness of the two loyalty dimensions is examined in the social media (SM) context by testing if attitudinal SM E-loyalty is positively related to behavioral SM E-loyalty.

Furthermore, the construct social media engagement (SME) has gained importance within the social media domain. SME can also be viewed as a multidimensional construct consisting of a passive and an active dimension, which can be expressed in engagement behaviors on SM.

The different types of engagement are expected to exert differential effects on SM E-loyalty dimensions. This study examines to what extent E-commerce companies can develop behavioral SM E-loyalty through attitudinal SM E-loyalty, mediated by passive and active SME individually. An online questionnaire conducted among 218 respondents showed that behavioral SM E-loyalty consisted of two sub-dimensions, namely ‘social media behavior’

and ‘transactional behavior’. Our findings showed that attitudinal SM E-loyalty was positively related to the sub-dimension social media behavior, but it was not significantly related to the sub-dimension transactional behavior. Furthermore, both passive and active SME positively mediated the relationship between attitudinal SM E-loyalty and the sub- dimension social media behavior. However, only passive SME mediated the relationship between attitudinal SM E-loyalty and the sub-dimension transactional behavior, although the mediating effect turned out to be negative instead of positive. The findings of this study contribute to marketing literature by bridging the gap about the interrelatedness between these two loyalty dimensions within the SM context, which until now were solely studied

individually. Lastly, this study contributes to marketing literature by being the first study to provide evidence for the mediation and differentiating roles of passive and active SME on the relationship between attitudinal and behavioral SM E-loyalty.

Keywords: Attitudinal SM E-loyalty, Behavioral SM E-loyalty, Passive social media

engagement, Active social media engagement, Social media behavior, Transactional behavior

(3)

Table of content

1. Introduction ... 5

2. Literature review ... 8

2.1. Customer loyalty, E-loyalty and SM E-loyalty ... 8

2.1.1. Dimensions of SM E-loyalty ... 9

2.1.2. Attitudinal SM E-loyalty ... 9

2.1.3. Behavioral SM E-loyalty ... 11

2.1.4. Relationship between attitudinal and behavioral SM E-loyalty ... 12

2.2. Social media engagement ... 13

2.2.1. Relationships between passive and active social media engagement and attitudinal SM E-loyalty ... 14

2.2.2. Relationships between passive and active social media engagement and behavioral SM E-loyalty ... 16

2.3. Mediating role of passive and active social media engagement on the relationship between attitudinal SM E-loyalty and behavioral SM E-loyalty ... 17

2.4. Conceptual model and hypotheses ... 18

3. Methodology... 19

3.1 Research Design: Data collection and sampling strategy ... 19

3.2 Procedure and Questionnaire... 20

3.3. Principal component analysis ... 22

3.4. Assumptions check: Linearity, normality and independence ... 22

3.5. Analyzing methods ... 23

4. Results... 23

4.1. Data cleaning ... 23

4.2. Factor analyses ... 24

4.2.1. Initial exploration of theoretical constructs: Distinguishing two sub-dimensions of behavioral SM E-loyalty ... 24

4.2.2. Final factor analysis: Behavioral SM E-loyalty construct divided into ‘social media behavior’ and ‘transactional behavior’ ... 25

4.3. Reliability analysis ... 27

4.4. Assumptions check: Linearity, normality and independence ... 28

4.5. Mediation analyses ... 30

4.5.1. Mediating relationship between attitudinal SM E-loyalty, active SME and social media behavior ... 31

4.5.2. Mediating relationship between attitudinal SM E-loyalty, passive SME and social media behavior ... 32

4.5.3. Mediation relationship between attitudinal SM E-loyalty, active SME and transactional behavior... 32

(4)

4.5.4. Mediation relationship between attitudinal SM E-loyalty, passive SME and

transactional behavior... 33

5. Discussion ... 34

5.1. Main findings and theoretical implications ... 34

5.2. Managerial implications of our findings ... 37

5.3. Limitations of our findings ... 38

6. References ... 40

7. Appendix ... 46

Appendix A: Questionnaire ... 46

Appendix B: Initial factor analysis ... 56

Appendix C: Final constructs and items... 58

Appendix D: Assumptions check for multiple regression analysis ... 59

(5)

1. Introduction

Customer loyalty is a well-known topic in marketing literature, which is mainly due to “its importance in achieving sustainable competitive advantages and financial outcomes”

(Tartaglione, Cavacece, Russo & Granata, 2018, p. 1). Scholars often view customer loyalty as a multidimensional construct (e.g., Dick & Basu, 1994), consisting of an attitudinal and a behavioral dimension. Throughout the years customer loyalty has been applied in different contexts. According to multiple scholars the multidimensional view holds across the electrical context and the social media context as well (Donio, Massari, & Passiante, 2006; Lopez- Miguens & Vázquez, 2017; Toufaily, Ricard, & Perrien, 2016; Huang, 2017).

Within the electrical commerce context, customer loyalty is referred to as customer E-loyalty.

E-loyalty has gained a greater interest and importance among researchers over the years (López-Miguens & Vázquez, 2017; Toufaily et al., 2016). In more recent studies, the focus of E-commerce seems to have shifted into a new technological context, namely the social media context, where researchers refer to it as SM E-loyalty (van Asperen, Rooij, & Gijkmans, 2018; Salem, Tarofder, Chaichi, & Musah, 2019; Zhao, Chen, & Wang, 2016). The main distinguishing factor between SM E-loyalty compared to E-loyalty is the participation of the customer with the company’s SM (Alfonzan, de Groot, Shields, & Meeran, 2020).

In this study we are going to examine the relationship between attitudinal and behavioral SM E-loyalty. Prior research gives us reason to assume that behavioral SM E-loyalty originates from attitudinal SM E-loyalty. This assumption is already established in the original retail environment (e.g., Dick & Basu, 1994; Srivastava & Kaul, 2016) and within the internet (E- commerce) environment (e.g., Husain, 2017). This study will examine the same relationship in the social media environment. Understanding the interrelatedness between attitudinal and behavioral SM E-loyalty could provide important contributions for marketeers who seek to increase customers behavioral SM E-loyalty through attitudinal SM E-loyalty. Therefore we will investigate the interrelationship of these two constructs. Our findings could further validate the limited insights that exist regarding the interrelatedness of attitudinal and

behavioral loyalty in the SM context from scholars in the field (Husain, 2017; Kang, Tang, &

Lee, 2015; Mainardes, & Cardoso, 2019).

(6)

Within this study we will relate the SM E-loyalty field of research to another field of research, namely social media engagement (SME). SME is the ‘engagement’ of the customer with the SM platform, sometimes simply referred to as social media use (van Asperen et al., 2018;

Zhao et al., 2016). As of yet, there is still a lot unknown about the effects of social media use (SME) on customer loyalty (van Asperen et al., 2018). A variety of studies have shown that a stronger SME can result in increased SM E-loyalty (van Asperen et al., 2018; Hardeep, Wirtz,

& Anu 2020; Salem et al., 2019). However, most of these studies examined the relationship between SME on a single dimension of SM E-loyalty only, where they either focused on attitudinal (e.g., van Asperen et al., 2018; Salem et al., 2019) or the behavioral dimension (e.g., Yoshida, Gordon, Nakazawa, Shibuya, & Fujiwara, 2018) of SM E-loyalty. The present study will investigate how SME relates to both attitudinal and behavioral SM E-loyalty.

Most researchers have acknowledged the multidimensionality of SM E-loyalty (Lopez- Miguens & Vázquez, 2017; Toufaily et al., 2016; Donio et al., 2006; Huang et al., 2017) but an emerging number of scholars also acknowledge the multidimensionality of the construct SME (van Asperen et al., 2018; Khan, 2017; Men & Tsai, 2013; Yu, 2016). According to these scholars SME includes a passive (consuming or lurking) and an active (contributing) engagement dimension. Earlier studies on SME did not distinguish between these two dimensions (Brodie, Hollebeek, Juric, & Ilic , 2011; Calder & Malthouse, 2008; Plummer, 2006). However, passive and active SME differ considerably in their degree of interactivity and therefore, the different engagement types can lead to different outcomes (Yu, 2016).

Within the SM E-loyalty context there seems to be some initial evidence that passive and active SME relate differently to attitudinal SM E-loyalty (van Asperen et al., 2018). However, van Asperen and colleagues (2018) mentioned that the relationship between SME and

attitudinal SM E-loyalty and SME might be reversed. Attitudinal SM E-loyalty might

positively influence SME as loyal customers might engage stronger with a company’s SM. As of yet, no studies have investigated how passive and active SME are individually related to behavioral SM E-loyalty. Based on the findings of (van Asperen et al., 2018) we expect active and passive SME to affect behavioral SM E-loyalty differently.

The present study examines the relationship between active versus passive SME and both attitudinal loyalty and behavioral loyalty in the SM context. More specifically, we will explore the mediating roles of passive and active SME on the relationship between attitudinal and behavioral SM E-loyalty. By doing so, we try to integrate the research fields focusing on

(7)

passive and active SME with the field of attitudinal (e.g., van Asperen et al., 2018; Salem et al., 2019; Salem & Salem, 2019) and behavioral SM E-loyalty (e.g., Hardeep et al., 2020;

Yoshida et al., 2018). An integration of these independent yet overlapping fields of research may be beneficial in order to validate the assumption that different SME types could have differential associations with SM E-loyalty dimensions. Understanding the relationship between attitudinal SM E-loyalty and active and passive SME and the way in which they subsequently explain behavioral SM E-loyalty will help to get a better understanding of the SME construct and how it can be deployed to benefit SM E-loyalty marketing goals.

The research question that this study aims to answer is: To what extent can E-commerce companies develop behavioral SM E-loyalty through attitudinal SM E-loyalty, mediated by passive vs active social media engagement?

(8)

2. Literature review

2.1. Customer loyalty, E-loyalty and SM E-loyalty

Customer loyalty in its retail origin is focused on the ‘loyalty relationship’ between an

individual’s relative attitude towards a company and the repeat (purchase) patronage (Dick &

Basu, 1994). With customer loyalty, companies can achieve sustainable competitive advantages that could ultimately result in positive financial outcomes (Tartaglione et al., 2018). Gaining financial benefits makes the construct customer loyalty commercially relevant for companies that seek to exploit marketing activities. The commercial relevance makes customer loyalty a popular marketing topic among scholars in marketing literature (Tartaglione et al., 2018).

The construct of customer loyalty can be applied in the online context or rather the internet environment, where it is referred to as E-loyalty. The theoretical foundations of E-loyalty are similar to retail customer loyalty (Alfonzan et al., 2020). The main difference between E- loyalty compared to traditional loyalty is trust and information asymmetry (Alfonzan et al., 2020). Customers could be less likely to trust a company that they only know from the online environment in contrast to an offline store where there is much less information asymmetry.

In more recent studies, the focus of E-loyalty seems to have shifted into a new technological context, namely the social media context, where researchers refer to it as SM E-loyalty (van Asperen et al., 2018; Salem et al., 2019; Zhao et al., 2016). The main difference between E- loyalty and SM E-loyalty is ‘participation’. Participation is referred to as active contributing behavior (e.g., sharing, posting comments) on a company’s SM platform. Through

participation customers add to the content of the platform, which makes SM E-loyalty significantly different from E-loyalty. The added content created by participation in a company’s SM platform may create more awareness (Alfonzan et al., 2020) and more favorable perceptions of the company’s products and services among its customers (Davis, 2010). Creating more awareness is the most acknowledged benefit of user participation (Davis, 2010). Increasing awareness is commercially interesting, but what makes SM

especially interesting compared to the general internet environment is that SM also increases traffic to a company’s website (Davis, 2010), which shows that the benefits of the SM environment reach beyond its own environment and also impact the company’s internet environment. SM E-loyalty will provide more commercial advantages compared to loyalty in

(9)

general, or E-loyalty outside the SM context. Therefore, this study will focus on SM E-loyalty specifically.

2.1.1. Dimensions of SM E-loyalty

Dick & Basu’s (1994) conceptualization of customer loyalty acknowledges that it includes both an attitudinal and a behavioral dimension. Similar to traditional customer loyalty, the multidimensional view towards the construct seems to hold merit across different contexts, including the electrical (e.g., Lopez-Miguens & Vázquez, 2017; Toufaily et al., 2016) and the SM context (e.g., Tseng, Cheng, Li, & Teng, 2017; Jodl, 2013). Hence, within this study we will view SM E-loyalty as a multidimensional construct.

The attitudinal loyalty dimension can be expressed through customer attitudes, preferences and purchase intentions, that is, excluding actual purchase behaviors (Nam, Ekinci, & Whyatt, 2011). The behavioral dimension aims at actual behaviors and is expressed through

purchasing/visiting (or re-purchasing/re-visiting) behaviors and participation behaviors to the company’s SM platform (Alfonzan et al., 2020; Husain, 2017; Nam et al., 2011; Toufaily et al., 2016).

2.1.2. Attitudinal SM E-loyalty

Attitudinal loyalty finds it origin at the customer’s attitude towards a company (or its

services/products). Salem & Salem (2019) state that the construct of attitudinal loyalty should consist of attitudinal commitment (affective loyalty) and ultimately the intention to purchase from the company (conative loyalty). Although attitudes and intentions are different and independent constructs, the theory of planned behavior (Ajzen, 1991) links the two construct as an indirect explanatory mechanism for actual behaviors. We will take the theory of planned behavior as a point of departure. This same mechanism of attitudes (arising from affective loyalty) predicting intentions (arising from conative loyalty) is endorsed by (Salem et al., 2019 in the SM context. Furthermore, many scholars argue that affective loyalty and conative loyalty together shape the overarching construct attitudinal loyalty (e.g., van Asperen et al., 2018; Russel-Bennett, McColl-Kenedy, & Coote, 2007; Evanschitzky & Wunderlich, 2006).

Hence, these scholars link both the attitudinal and intentional component of Ajzen’s theory into one construct, where affective loyalty is viewed as the first stage affecting the second stage conative loyalty. Finally, Ajzen (1991) explains that the stronger one’s intentions, the

(10)

more likely one will perform this intended behavior. This completes the mechanism starting with attitude towards the behavior (affective loyalty), to behavioral intentions (conative loyalty) and ultimately actual customer behavior (behavioral loyalty).

The affective part of attitudinal SM E-loyalty can be described as affective feelings or rather a

‘psychological bond’ between the customer and the company (van Asperen et al., 2018).

Within the social media context these affective feelings can be expressed in the attachment of a customer towards the company (van Asperen et al., 2018). The conative part of attitudinal SM E-loyalty refers to a customer’s willingness to take action or rather a “customers constant behavioral intention towards a particular company” (Salem et al., 2019, p. 384). Within the social media context behavioral intentions (conative) can be expressed through the

willingness or dedication to acquire a company related product/service (Salem et al., 2019;

Salem & Salem, 2019). Toufaily and colleagues (2016) define attitudinal SM E-loyalty in their meta-analysis in a more inclusive way. They mention that attitudinal loyalty is mainly determined by intentions of repeat behavior. These intentions can be related to multiple loyalty expressions, such as intentions to revisit the platform, to make a purchase, to maintain relationship or to recommend the platform, not just intentions related to the purchase act. The definition of attitudinal SM E-loyalty used in this study is based on definitions of (Salem et al., 2019; Toufaily et al., 2016) as these combined definitions describe the construct in an inclusive way, taking both affective and conative parts into account: Attitudinal SM E-loyalty is a customer’s constant behavioral intention (not actual behavior) of repeat behaviors related to purchasing, visiting, recommending and maintaining a relationship with the company.

(11)

2.1.3. Behavioral SM E-loyalty

Behavioral loyalty is referred to as a behavior that consists of an individual’s repeated

purchases of a company’s products/services. This (re)purchasing behavior is the key behavior that should be included within the behavioral dimension according to (Husain, 2017; Nam et al., 2011). Unlike the attitudinal dimension of SM E-loyalty, there seems to be more

agreement on which behaviors should be included in the behavioral dimension of SM E- loyalty. (Re)purchasing behavior also includes increased share of wallet according to (Jones

& Taylor, 2007; Toufaily et al., 2016) as loyal customers tend to allocate increasingly more money towards the company. Furthermore, the customers (re)visiting behavior is considered as an important sub-dimension of behavioral SM E-loyalty, akin to (re)purchasing behavior according to (Nam et al., 2011; Toufaily et al., 2016). As it shows a customer’s loyalty in a behavioral manner.

Behavioral loyalty also has a number of sub-dimensions, which are not directly related to the customer’s act of purchase. Firstly, the most distinctive behavioral expression of behavioral SM E-loyalty compared to E-loyalty is participation behaviors to the company’s SM platform, through which participants add to the SM content of the company (Alfonzan et al., 2020).

Some studies include psychological sub-dimensions to behavioral loyalty as well (e.g., Jones

& Taylor, 2007; Toufaily et al., 2016). They include non-behavioral aspects like purchase intentions and switching intentions, which are closer related to conative loyalty. Therefore, we will not include behavioral intentions as aspects of behavioral loyalty. Lastly, Toufaily and colleagues (2016) also include WOM as an expression of behavioral loyalty. We will exclude WOM as an expression of behavioral loyalty because we view that it is rather an active type of SME instead, which we will relate to behavioral SM E-loyalty in our conceptual model.

To distinguish clearly between attitudinal and behavioral SM E-loyalty, in this study behavioral loyalty will be solely focused on customer’s self-reported behaviors (not

intentions) on a company’s SM platform. The definition of behavioral SM E-loyalty used in this study is based on definitions of (Alfonzan et al., 2020; Husain, 2017; Jones & Taylor, 2007; Nam et al., 2011; Toufaily et al., 2016), that is: “the constant behavioral commitment of

(12)

a customer with regards to repeat behaviors of purchases of online products and services, visiting of the company’s SM platform and the participation behaviors on the platform”

2.1.4. Relationship between attitudinal and behavioral SM E-loyalty

Previous research has established that attitudes are important determinants of behavior (Dick

& Basu, 1994), which is in line with the theory of planned behavior of Ajzen (1991). First of all, Ajzen (1991) established with high accuracy that attitudes towards the behavior can predict an individual’s behavioral intentions. If we relate this theory to our study context we could assume that attitudes towards the company (affective loyalty) could translate into behavioral intentions (conative loyalty). Affective and conative loyalty are the sub- dimensions which together shape attitudinal loyalty. Ajzen subsequently states that “the stronger the intention to engage in a behavior, the more likely should be its

performance”(Ajzen, 1991, p. 181). Hence, we assume that the behavioral intentions that arise from attitudinal SM E-loyalty could translate into behavioral SM E-loyalty.

Several studies have examined the relationship between attitudinal and behavioral loyalty (e.g., Husain, 2017; Srivastava & Kaul, 2016). Within the retail context Srivastava & Kaul proved that there is a positive and direct effect of attitudinal loyalty on behavioral loyalty (2017). Also in the context of E-commerce (Husain, 2017) showed that attitudinal E-loyalty was positively related to behavioral E-loyalty. These findings seem to be in line with the popular theory of planned behavior of Ajzen (1991).

As of yet, no studies have validated the assumptions regarding the relationship between attitudinal loyalty and behavioral loyalty within the SM context. Validating assumptions about the interrelatedness between these two loyalty dimensions within the SM context would bridge the gap in SM E-loyalty literature where these constructs are solely studied

individually. These findings could also be commercially interesting for companies that seek to increase the behavioral SM E-loyalty of their customers. We therefore hypothesize that:

H1. There is a positive relationship between attitudinal SM E-loyalty and behavioral SM E-loyalty.

(13)

2.2. Social media engagement

Nowadays it is almost unavoidable for companies to own and use social media (SM)

platforms for commercial purposes. SM has been used with the main purpose to engage and connect with customers. Furthermore, social media engagement (SME) is assumed to be strategically valuable to further improve company performances (van Asperen et al., 2018).

More specifically, SM platforms are regarded as attractive tools to develop SM E-loyalty among customers towards the company (van Asperen et al., 2018).

Within the SM context there is no clear agreement among scholars about the definition of SME. However, the definition of customer engagement by Plummer (2006) can be applied in multiple contexts, including the SM context. Plummer’s definition suggests that SME finds it origin from a company’s communication, transferred by a communication channel (e.g., SM) towards its customers, and the engagement of customers with the company’s platform can be expressed through actions/behaviors (e.g., comments, sharing, co-creating).

The way customers communicate or ‘engage’ with a company’s SM channel is studied more extensively by Brodie and colleagues (2011) in an exploratory study on customer engagement within the online brand community context. The online brand community context is closely related to the SM context because of its social nature and interactive possibilities among participants. While Plummer suggests that SME includes only communication with the company’s platform. Brodie and colleagues (2011) extend this definition in their exploratory study. They found that next to communication with the company’s SM platform, SME also includes interactive experiences with other members of the platform. Calder & Malthouse (2008) also acknowledge that interactive experiences are an important part of SME (they refer to ‘media engagement’ instead of SME) as well. However, Calder & Malthouse (2008) extend the definition of SME even further by including co-created experiences as well. These two engagement sub-dimensions seem to distinguish SME from online brand engagement, making it a more social construct.

(14)

The definition of SME in the present study acknowledges that both interactivity with the company’s SM platform and with other members of the platform and co-creation are key aspects of SME and we therefore define SME as: Social media engagement consists of the interactive and/or co-created communication experiences of members of the SM platform with the company or other members of the SM platform.

2.2.1. Relationships between passive and active social media engagement and attitudinal SM E-loyalty

Social media engagement (SME) can be expressed in many different ways (e.g., reading company posts, liking, sharing etc.). Most scholars tend to categorize SME in passive versus active engagement activities (e.g., van Asperen et al., 2018; Dessart, Veloutsou, & Thomas, 2015; Men & Tsai, 2013). For example, Dessart and colleagues (2015) studied the dimensions of SME in online brand communities and distinguished passive SME, including passive expressions such as reading and consuming SM content and investing time on the SM platform from active SME, which included active expressions such as sharing, co-creating and advocating.

As the two engagement types differ considerably in their degree of interactivity (consuming vs sharing) and co-creation (which is primarily an active SME expression), it seems likely that they contribute differently to specific outcome variables as well. In other research

contexts (e.g., the psychological context) it is already established that passive and active SME exert differential effects on users well-being (Verduyn et al., 2015) and loneliness (Deters &

Mehl, 2012).

There is some initial evidence that passive and active SME contribute to the explanatory power of SM E-loyalty in a unique way as well. That is, van Asperen and Colleagues (2018) found that passive SME positively affected sub-dimensions of attitudinal SM E-loyalty, while the positive effect of active SME on sub-dimensions of attitudinal SM E-loyalty were non- significant. Salem and colleagues (2019) provided further evidence for this relationship. They found that social media activities (similar to SME), including both passive and active

expressions affects affective- and subsequently also affects conative loyalty, which together comprise attitudinal loyalty. The studies of van Asperen et al., (2018) and Salem et al., (2019) raise the assumption that both types of SME can affect attitudinal SM E-loyalty.

(15)

To the author’s knowledge, there are only two studies that have examined how passive and active SME affect attitudinal SM E-loyalty (i.e., van Asperen et al., 2018; Salem et al., 2019).

However, these findings need to be extended in three important ways.

First, both studies that have been done in relation to attitudinal SM E-loyalty assume that (active and passive) SME affect attitudinal SM E-loyalty rather than vice versa. This is often reasoned because following or becoming actively engaged to a company’s SM platform could yield positive perceptions, which lead to loyalty (van Asperen et al., 2018; Salem et al., 2019). However, an alternative reasoning could be that loyal customers might engage stronger (e.g., consume more, contribute more) with the company’s SM platform, compared to non- loyal customers because they want to maintain their relationship with the company. This assumption indicates the reversed relationship. Indeed, van Asperen and colleagues (2018) acknowledged that the relationship between SME and attitudinal SM E-loyalty and SME might be reversed (2018). Therefore, our study will examine the reversed relationship by testing to what extend attitudinal SM E-loyalty affects both passive and active SME individually.

Second, although van Asperen et al., (2018) showed that active and passive SME relate differently to attitudinal SM E-loyalty, their results specifically show that active SME relates weaker and even non-significant to attitudinal SM E-loyalty dimensions while passive SME is significantly related. This finding indicates that passive SME might be a stronger predictor for attitudinal SM E-loyalty compared to active SME.

Third, the studies that have been done are performed in specific contexts only, namely van Asperen et al., (2018) studied the constructs in the travel context and Salem et al., (2019) in the fast fashion context. The loyalty of customers of a travel agency could differ considerably to other contexts. The extent to which attitudinal SM E-loyalty and passive and active SME are related could be context dependent. This gives us reason to assume that (contrary to findings of van Asperen et al., 2018), attitudinal SM E-loyalty could be positively related to active SME as well, but likely to a lesser extent, based on the findings of (van Asperen et al., 2018). Therefore, our study will test if attitudinal SM E-loyalty is stronger related to passive SME than to active SME. We therefore hypothesize that:

(16)

H2A: Attitudinal SM E-loyalty is positively related to passive SME H2B: Attitudinal SM E-loyalty is positively related to active SME

H2C: Attitudinal SM E-loyalty is more strongly related to passive than to active SME

2.2.2. Relationships between passive and active social media engagement and behavioral SM E-loyalty

Social media engagement (SME) can explain customer intentions arising from attitudinal SM E-loyalty as was assumed by (e.g., van Asperen et al., 2018; Salem et al., 2019). However, according to several scholars (e.g., Chahal, Wirtz, & Verma, 2020; Yoshida et al., 2018), SME can also explain customer behaviors arising from behavioral SM E-loyalty. In research focusing on the relationship between SME and behavioral SM E-loyalty, SME is studied as a whole (e.g., Chahal et al., 2020), and as a multidimensional construct but only focusing on one dimension (e.g., Algharabat, Rana, Alalwan, Baabdullah, & Gupta, 2020; Yoshida et al., 2018).

Multiple scholars have examined how passive and active SME affect behavioral loyalty, but each scholar took a slightly different approach with regards to the constructs. However, contrary the findings of (van Asperen et al., 2018) in relation to attitudinal SM E-loyalty, there is no clear initial evidence that indicates that passive and active SME contribute to the explanatory power of behavioral SM E-loyalty in a unique way as well.

SME as a whole (including both passive and active expressions) was found to be positively related to brand loyalty according to (Chahal et al., 2020). Chahal and colleagues did not distinguish between passive and active SME, but their findings indicates that both passive and active SME could be positively related to behavioral SM E-loyalty. Therefore, our study will be the first to examine to what extend passive and active SME affects behavioral SM E- loyalty individually.

Other scholars (e.g., Yoshida et al., 2018) acknowledge the multidimensionality of ‘consumer SM brand engagement’ (similar to SME) and tested how active SM brand engagement

(similar to active SME) related towards behavioral loyalty outcomes. They established a positive relation between active SME and the behavioral brand loyalty dimension. These findings indicate that active SME can affect SM E-loyalty, so customers who are actively

(17)

engaged to the company’s SM tend to be more loyal compared to customers who are not actively engaged to the company’s SM. Algharabat and colleagues (2020) provide further evidence for this relationship. They studied the construct ‘customer participation to the company’s social media site’, which is similar to active SME. Customer participation had a positive effect on brand loyalty (including behavioral components) towards the social media page of the company according to (Algharabat et al., 2020).

The findings of Chahal et al., (2020) raise the assumption that both passive and active SME are positively related to behavioral SM E-loyalty. We will firstly test if passive and active SME contribute to the explanatory power of behavioral SM E-loyalty uniquely to get a better understanding of the individual interrelations. Furthermore, the studies of Algharabat et al., (2020) and Yoshida et al. (2018) are strongly focused on active SME in relation to behavioral SM E-loyalty. Their findings raise the assumption that active SME might be a stronger predictor for behavioral SM E-loyalty compared to passive SME. Establishing differential effect of passive and active SME would be a valuable contribution to marketing literature as it will give marketeers a better understanding of the individual dimensions in relation to

behavioral SM E-loyalty. This will also be the first study that examines both engagement types in relation behavioral SM E-loyalty. Hence, we hypothesize that:

H3A: Passive SME is positively related to behavioral SM E-loyalty H3B: Active SME is positively related to behavioral SM E-loyalty

H3C: Active SME is more strongly related to behavioral SM E-loyalty than passive SME

2.3. Mediating role of passive and active social media engagement on the relationship between attitudinal SM E-loyalty and behavioral SM E-loyalty

This study will firstly contribute to marketing literature by validating the limited findings that exist about the interrelationships regarding SM E-loyalty dimensions to bridge the gap within SM E-loyalty literature about the interrelatedness of the two dimension.

Secondly, multiple scholars (e.g., Algharabat et al., 2020; Chahal et al., 2020; Yoshida et al., 2018) established that customers who are more engaged (both actively and passively) with a company’s SM tend to be more behaviorally loyal compared to customers who are not

(18)

engaged to the company’s SM. These findings have never been validated within the SM context.

Furthermore, the main purpose of this study is to integrate the social media engagement (SME) field of research to the SM E-loyalty field of research by testing if the relationship between attitudinal and behavioral SM E-loyalty operates indirectly through passive and active SME. More specifically, we expect active and passive SME to be individual mediators on the relationship between attitudinal and behavioral SM E-loyalty. We expect customers with positive behavioral intentions (attitudinal SM E-loyalty) to be more engaged (both passively and actively) towards a company’s SM platform, which will subsequently affect their behavioral SM E-loyalty. Our study aims to create a better understanding of the individual explanatory function of SME dimensions by examining their mediating roles within the SM E-loyalty mechanism. Understanding how the individual SME dimensions could be deployed to increase behavioral SM E-loyalty could be financially valuable to marketeers. Hence, we hypothesize:

H4A: Passive social media engagement will positively mediate the relationship between attitudinal SM E-loyalty and behavioral SM E-loyalty

H4B: Active social media engagement will positively mediate the relationship between attitudinal SM E-loyalty and behavioral SM E-loyalty

2.4. Conceptual model and hypotheses

Figure 1 displays the conceptual model of this study with our hypotheses included.

Figure 1. Conceptual model

(19)

3. Methodology

3.1 Research Design: Data collection and sampling strategy

In order to obtain useful insight on how different types of SME affect SM E-loyalty dimensions the cross-sectional research design was employed by conducting an online quantitative survey in Qualtrics (“Online Survey Software – Trusted by +5.5M Survey

Creators | Qualtrics UK”, 2019). Quantitative research is an appropriate way to quantify large amounts of statistical data and generalize the results from the sample to the population of interest (Malhotra, 2010). Using the survey software Qualtrics is a convenient way to obtain primary data and subsequently analyze the data because the software makes it possible to import data to analyzing software like SPSS.

The target population consisted of individuals who had made online purchases at least twice within the last 12 months. Furthermore, these individuals had to follow the social media channel of the company of which they purchased online from. These two sampling criteria ensured that all participants could be tested on the hypotheses proposed in chapter 2. We did not distinguish between specific social media platforms as sampling criteria because this study investigates SM E-loyalty, taking the whole SM context into account. Participants did have to indicate their most frequently used SM platform, because this could yield useful insights for managerial implications.

The survey was distributed among people who are likely to purchase online and follow the SM platform of the company they buy from, mainly through email, Facebook groups and WhatsApp. Social media is a convenient tool to acquire many participants in an efficient manner (Blumberg, Cooper, & Schindler, 2008). The sampling method that was used to obtain the research data was convenience sampling. This study has only few sampling criteria and the two filter questions filter out respondents who are not useable, which results in a broad and large target population. Convenience sampling is based on the ease of access of participants and it is viewed as an appropriate method to target such a broad population conveniently (Blumberg et al., 2008; Malhotra, 2010).

To ensure enough statistical power, the ‘rules of thumb’ of Voorhis & Morgan (2007) were used to determine our minimum sample size. For the purpose of a factor analysis, which we

(20)

are going to perform within this study, a total of ~200 respondents can be viewed as ‘fair’,

~300 as ‘good’, ~500 as a ’very good’ and ~1.000 as an excellent amount of respondents (Tabachnick & Fidell, 1996). ‘50 respondents per factor’ is another rule of thumb that is often used for the purpose of a factor analysis (Pedhazur & Schmelkin, 1991), which would indicate that we need a minimum of 200 respondents as we have four factors within our conceptual model. Hence, a minimum of 200 respondents seems to be sufficient to test our hypotheses.

3.2 Procedure and Questionnaire

The questionnaire started with 2 filter questions in which participants had to indicate (1) that they made an online purchase at least twice within the last 12 months and (2) that they follow the company that they bought from on a SM platform. The 2 filter questions ensured that our sampling criteria were met. After that, participants had to fill in the name of the company, which would then be inserted into the statement questions to make the questionnaire more personally relevant. Furthermore, the questionnaire continued with multiple statements on which the participants had to indicate their level of agreement or the frequency of which they perform a certain proposed behavior related to passive and active SME, attitudinal and behavioral SM E-loyalty and some additional confounding variables.

The questions of the questionnaire were based on validated scale items from previous studies regarding the constructs that are incorporated into our study (see Appendix A for an overview of all items of the questionnaire). All items related to the main constructs were answered on a 7-point Likert scale, either ranging from 1 (strongly agree) to 7 (strongly disagree) or ranging from 1 (never) to 7 (always).

Attitudinal SM E-loyalty and behavioral SM E-loyalty.

To measure our independent and dependent variables, we adapted the scale items of (Alfonzan et al., 2020) which were obtained from their literature review. The items from (Alfonzan et al., 2020) are suitable for our study, because they are specifically focused on

‘social media’ loyalty dimensions rather than traditional or E-loyalty. The results of the factor analysis of Alfonzan and colleagues (2020) showed that there are seven dimensions of SM E-loyalty of which three could be related to attitudinal SM E-loyalty (contextual

intentions, participation intentions and transactional intentions) and three could be related to behavioral SM E-loyalty (Participation behavior, contextual behavior and transactional

(21)

behavior). One dimension, namely company preference could be viewed as a separate category and is therefore excluded from this study.

We included 10 attitudinal SM E-loyalty items, which were all statement questions which needed to be answered on a 7-point Likert scale ranging from 1 (strongly agree) to 7 (strongly disagree) (e.g., The probability that I will encourage friends and others to buy from the

website of [Company] is high). Furthermore, we included 9 behavioral SM E-loyalty items, of which 7 questions were statement questions which needed to be answered on a 7-point Likert scale ranging from 1 (strongly agree) to 7 (strongly disagree), similar to the attitudinal

questions. Additionally, we added 2 open ended interval questions regarding respondents transactional behavior, because ordinal scale questions were less suitable to obtain

respondents wide ranging answers (e.g., What is the amount of money you have spent with this company in euro’s? (approximately).

Social Media Engagement (SME).

To measure our independent variables passive and active SME, the majority of the scale items were adopted from van Asperen et al., (2018), who based them on the instrument of Men &

Tsai (2013). We included 5 passive SME items and 4 active SME items, which were all statement questions which needed to be answered on a 7-point Likert scale ranging from 1 (never) to 7 (always) (e.g., I ‘like’ posts on the social media page of [Company]. The questions related to the construct SME (both passive and active SME) were slightly adapted to fit the context of our study, but the content of the questions remained the same to ensure their validity. To measure passive SME, we added the ‘learning’ sub-dimension of Dessart et al., (2015) (see also Chahal et al., 2020) because this dimension is expressed through passive engagement expressions (e.g., reading posts or seeking for specific information).

Confounding variables.

The 5 confounding variables that we included into our study are online shopping experience, social media experience, satisfaction, trust and perceived quality. Online shopping and social media experience were included to account for respondent competences regarding the

technology involved within our study. Satisfaction and trust were mentioned as the two most studied determinants of online customer loyalty by the literature review of (Toufaily et al., 2016). Perceived quality is positively related to brand loyalty within the SM context

according to (Shanahan, Tran, & Taylor, 2019). Quality is also featured as a determinant of

(22)

online customer loyalty by (Toufaily et al., 2016). The confounding variable questions were all statement questions which needed to be answered on a 7-point Likert scale (e.g., Rate the level of quality of the products of [Company]). Lastly, some general demographic control variables were included, namely age, gender, educational level, work status and income to ensure the validity of our outcomes.

3.3. Principal component analysis

In order to ensure the validity and the reliability of our constructs we used a factor analysis (FA) to validate if the expected dimensions related to our constructs also exist in our dataset.

Before we started our factor analysis, we had to establish if it is appropriate to perform a factor analysis by using multiple methods. The first method that we used was the Kaiser- Meyer-Olkin (KMO) measure of sampling adequacy, to establish correlation and covariance between variables. Variables should be >.50, variables <.50 should be left out of the analysis (Malhotra, 2010). Next Bartlett’s test of sphericity was performed to confirm that the

variables correlate with each other and are appropriate for factor analysis. Therefore the null hypothesis of Bartlett’s test should be rejected (Bartlett, 1950). Lastly, we checked for communalities using the principal component analysis (PCA). The amount of variance that is explained by our factors should be >.40 (Malhotra, 2010). After we established that FA was appropriate, we could start with the analysis.

3.4. Assumptions check: Linearity, normality and independence

In order to perform a multiple linear regression analysis we first need to perform assumption checks regarding the assumptions of linearity, normality and independence (Newbold, 2013).

For the linearity assumption we looked at the scatterplot of the relationship between our dependent variable and our independent variables and the extent to which it deviates from linearity. Furthermore, we want to establish that our data is normally distributed. To establish normality, we use a Normal P-P-Plot and a histogram to check the normality assumption.

Lastly, to check the independence assumption we want to establish no auto-correlation in our data. Therefore, we use the Durbin-Watson test to examine if our residuals are independent from each other. A Durbin-Watson test value between 1.5 < d < 2.5, indicates that residuals are indeed independent and that there is no auto-correlation (Newbold, 2013). The output of the statistical tests that were used to check the assumptions can be seen in Appendix D.

(23)

3.5. Analyzing methods

We use a cross-sectional mediation method to test our hypotheses. This means that our analysis assesses mediation using variables collected at a single point in time (O’Laughlin, Martin, & Ferrer, 2018). Because we want to predict the value of our dependent variable based on the value of our 3 independent variables, we use multiple regression as this is the appropriate statistical test to answer our hypotheses. For our multiple regression analysis, we use Hayes’ PROCESS macro for SPSS model 4 (Hayes, 2017). In order to establish a

mediator effect four critical conditions must be met according to Barron & Kenny (1986) to establish either a full mediation or partial mediation (or none). (1) The effect of the

independent variable on the dependent variable has to be significant (C-path = Total effect).

(2) The effect of the independent variable on the mediator has to be significant (a-path). (3) The effect of the mediator on the dependent variable has to be significant (b-path). The addition of the mediator has to make the remaining effect of the independent variable on the dependent variable (c’-path) non-significant (full mediation) or its effect should be lower than the total effect (partial mediation).

4. Results

4.1. Data cleaning

After the Qualtrics survey was closed we obtained a dataset with data from 439 respondents.

The data had to be cleaned, because about 50% of the respondents were not useable. Most of the respondents that were not useable did not get past the two filter questions. The first filter question was: (F1: Have you made online purchases at least twice in the last 12 months?), and the second filter question was (F2: Do you follow the social media of the company that you purchase online from?). The two filter questions showed that 211 respondents did not get past the second filter question. Therefore, these respondents could be removed from the sample. Additionally, 10 respondents filled in highly unrealistic values at the open-ended questions regarding the money spend at the company and the frequency of purchase at the company. These values could be viewed as outliers and therefore these respondents were excluded from the sample. After the data cleaning we proceeded with 218 useable respondents and reached our sample goal of a minimum of 200 respondents.

(24)

4.2. Factor analyses

4.2.1. Initial exploration of theoretical constructs: Distinguishing two sub-dimensions of behavioral SM E-loyalty

An initial exploration with an exploratory factor analysis (FA) resulted in the exclusion of multiple items that were originally assigned to the constructs attitudinal (AEL) and behavioral SM E-loyalty (BEL). This was expected because there are so many competing views among scholars regarding the definitions of attitudinal and behavioral SM E-loyalty. We had to solve these issues prior to conducting the final FA.

First, five out of ten items that were originally assigned to the attitudinal SM E-loyalty construct were deleted from the final analysis. More specifically, two attitudinal items (AEL7: I am interested in forwarding the message from the social media of [Company] and AEL8: When I post a comment about [QID1-ChoiceTextEntryValue], I will spend much effort) were excluded from the analysis because of high cross loadings on all other items. The attitudinal SM E-loyalty item AEL6 (I will comment on the messages from [Company] on its social media) loaded higher on the behavioral SM E-loyalty factor. As this item includes an expression of behavioral intentions rather than actual behavior we excluded the item from construct as well. Additionally, 2 items (AEL9: I will always buy from [Company], even if other competitors have slightly better offers. And AEL10: I consider buying from [Company]

rather than from other competitors, even if its prices are somewhat higher than others) loaded high on an independent factor, which could be labeled as ‘resistance to switch’ rather than attitudinal SM E-loyalty.

Second, based on the exploratory FA, there were also some amendments of the construct behavioral SM E-loyalty. Two of the seven items (BEL6: I have found occasions to mention the name of [Company] to others and BEL7: I was proud to tell others that I had bought from [Company]) did not load high on the same factor as our other behavioral SM E-loyalty items.

Instead, they loaded high on an individual factor, which could be labeled as positive word-of- mouth (WOM) rather than behavioral SM E-loyalty. We did not conceptualize positive WOM as part of behavioral SM E-loyalty in our literature review and the findings of our FA

supported this assumption. Hence, we excluded these 2 items from the final construct.

(25)

Furthermore, two other items of the behavioral dimension of SM E-loyalty (BEL8: What is the amount of money you have spent with this company in euro’s? (approximately) and BEL9:

How many times have you purchased from this company?) loaded high on the same single individual factor. As these two items could be interpreted as ‘transactional behavior’, which is a key subdimension to behavioral SM E-loyalty according to (e.g., Alfonzan et al., 2020;

Husain, 2017; Jones & Taylor, 2007; Nam et al., 2011), we decided to regard behavioral SM E-loyalty as a factor with two sub-dimensions, that is factor 1 “social media behavior”

(including 5 items: BEL1 to BEL5) and factor 2 “transactional behavior” (including 2 items:

BEL8 and BEL9). See appendix C for all constructs and corresponding items). All our analyses in relation to behavioral SM E-loyalty will be performed on these two separate sub- dimensions of behavioral SM E-loyalty. See Table 1 for the items that are included in the final theoretical constructs.

Lastly, our exploratory FA (see appendix D) showed that both active SME items and behavioral SM E-loyalty (social media behavior sub-dimension) items loaded high on the same individual factor. According to research (Algharabat et al., 2020; van Asperen et al., 2018; Yoshida et al., 2018) they are in fact independent constructs even though they can be strongly related to each other. Hence, we will view them as such and treat them as two independent constructs. Therefore, we will split up active SME and behavioral SM E-loyalty in our subsequent analyses. Results of a Pearson correlation analysis showed that the

constructs active SME and behavioral SM E-loyalty are indeed highly correlated to each other (.806). However, the items that belong to active SME correlate stronger with the construct active SME than with behavioral SM E-loyalty and vice versa. Hence, there seems to be no reason remove any items or to change the initial conceptual structure of our constructs due to collinearity issues.

4.2.2. Final factor analysis: Behavioral SM E-loyalty construct divided into ‘social media behavior’ and ‘transactional behavior’

In this section, we report the final factor analyses (FA) taking into account the decisions we have made above based on the exploratory analyses. In order to ensure that factor analysis (FA) is appropriate for our study the KMO and Bartlett’s test of Sphericity were performed on all items within the dataset. The KMO measure of sampling we obtained was .891, which can be viewed as a nearly ‘marvelous’ level of common variance (.90) according to (Kaisers &

Rice, 1974). Secondly, the Bartlett’s test of Sphericity was highly significant (p < .001), so

(26)

we can assume that the items correlate with each other and are appropriate for factor analysis.

Lastly, we checked for communalities using the principal component analysis (PCA). The amount of variance that is explained by each of our variables was >.40. All three tests established that FA is appropriate for our sample. The results of our factor analysis and reliability analysis can be viewed in Table 1.

Table 1

Factor analysis and Reliability Analysis

Constructs and items Factor

loadings

Commu- nalities

Eigenvalue/

Variance explained

Cronbach’s Alpha (α)

1. Passive SME 3.159 /

15.043

0.820 I view pictures on the social media

page of [Company]

.837 .725

I watch videos on the social media page of [Company]

.772 .683

I ‘like’ posts on the social media page

of [Company] .744 .653

I seek for specific content that I am interested in when looking at the social media page of [Company]

.613 .524

I read posts, user comments, or product reviews on the social media page of [Company]

.560 .410

2. Active SME 7.889 /

37.565%

0.930 I respond to comments made by other

followers of the social media page of [Company]

0.870 .801

I interact with other followers of the social media page of [Company]

0.843 .751 I share content of other followers

(e.g., share post or retweets) of the social media page of [Company]

0.838 .749

I share posts of the social media page of [Company] on my own social media page (e.g., video, audio, pictures, texts)

0.825 .724

3. Behavioral SM E-loyalty

3.1. Social media behavior 0.910

I have shared messages from the social media page of [Company]

0.880 .799 I have responded to comments of

other followers of the social media page of [Company]

0.878 .787

I have posted a comment on the social 0.783 .654

(27)

media page of [Company]

Many members of the social media page of [Company] used the

information that I recommend to them

0.765 .643

I have recommended a lot of information about [Company] to others through social media

0.660 .612

3.2. Transactional behavior 1.156 /

5.503

0.499 What is the amount of money

(approximately) you have spent with this company in euro’s in the last 12 months?

.846 .717

How many times have you purchased from this company in the last 12 months?

.757 .625

4. Attitudinal SM E-loyalty 1.641 /

7.815%

0.779 When I write about the experience

with [Company] on its social media, I will recommend it

.838 .703

I will recommend [Company] to someone who seeks my advice

.768 .661

The probability that I will encourage friends and others to buy from the website of [Company] is high

.689 .530

The probability that I would say positive things about [Company]

through social media (such as click

“like” in the fan page) is high

.660 .557

I would recommend the products of [Company] to my friends through social media

.645 .536

4.3. Reliability analysis

When we look at the internal consistency of our constructs we look for Cronbach’s alpha values greater than .60 as they can be considered acceptable (Malhotra, 2010). The Cronbach’s alpha values of the individual constructs active SME (.93) and social media behavior (SMB) (.91) are well above .60, which shows that the internal validity of our new separated constructs is better than when they were combined (.87) as can be seen in Table 1.

All other construct have a Cronbach’s alpha of >.60, except for our new construct

‘transactional behavior’ (TB), which has a Cronbach’s alpha value of .50. The Cronbach’s alpha measure is sensitive to the number of items in the scale and the items BEL8: What is the

(28)

amount of money you have spent with this company in euro’s? (approximately) and BEL9:

How many times have you purchased from this company? Use open-ended scales with high ranging values contrary to all other item scales, which are all 7-point noncomparative scales.

Furthermore, Cronbach’s alpha is known since inception to be very dependent on the number of items (TB: Only includes two items) and on the average inter-item correlation (Carmines &

Zeller, 1979). The inter-item correlation of TB is 0.33, compared to the inter-item correlation of SMB, which is 0.68. As a result Cronbach’s alpha tends to underestimate the internal consistency of the factor TB, which was indeed the case (α = .499). Therefore, the internal consistency of the construct TB can be estimated through structural equation modeling using the composite reliability, which assesses the internal validity while correcting for limiting underestimating limitations from Cronbach’s alpha. Indeed, composite reliability obtained a value of .75. This reliability is viewed as ‘satisfactory’ (Nunnally & Bernstein, 1994).

Within the FA, A VIF statistic indicates the strength of multicollinearity between constructs and hence, gives an indication whether the constructs are appropriate to use in multiple regression, because high multicollinearity can affect the results of our mediation analysis. All variables included in our study have low VIF-values of < 4 (see Table 2), which indicates that multicollinearity is not an issue among our construct according to (e.g.. Miles & Shevlin, 2001).

Table 2

Collinearity Statistics of all the IV’s

Constructs Tolerance VIF

Attitudinal SM E-loyalty .833 1.308

Behavioral SM E-loyalty

Social media behavior .314 3.184

Transactional behavior .941 1.063

Active SME .767 3.074

Passive SME .653 1.595

4.4. Assumptions check: Linearity, normality and independence

Most of our assumptions have been met as can be seen in Appendix D. The linearity of the relationship between the independent variables and transactional behavior (TB) is slightly better compared to the relationship with social media behavior (SMB) as a dependent variable. Furthermore, we can establish that our data is normally distributed. Both the P-P-

(29)

Plots generally seem to follow a straight line. Additionally, both the histograms show an approximately normal distribution. Lastly the Durbin-Watson test value for both of our models lie between 1.5 < d < 2.5, which indicates that there is no-auto correlation.

After checking for assumptions we can proceed with our multiple regression analyses. We want to perform 4 mediation analyses. We want to test the effect of attitudinal SM E-loyalty on each of our DV’s (behavioral SM E-loyalty and transactional behavior) individually, where the relationship is either mediated by active SME or by passive SME to examine how the strength of the mediators differentiate form each other.

Table 3 shows that both of our mediation models (with either SME or TB as a DV) are overall significant as they both have a significance levels <.05. Furthermore, our results show that the R2 for the mediation model with SMB as a DV is 68.6%, which is relatively high compared to the R2 for the mediation model with TB as a DV, which is only 5.9%. The individual R2

values indicate that a lot of the variance in SMB is explained by the IV’s and contrary, relatively little of the variance in TB is explained by the IV’s.

Table 3

R2 values of mediation models

Model indicated by DV R R2 F p Social media behavior .828 .686 155.767 .000 Transactional behavior .243 .059 4.486 .004

Furthermore, when we look at the individual contribution of our IV’s in explaining our DV’s we can see that the construct active SME does not significantly contribute to explaining TB (B = -.01, p = .84). Contrary, passive SME shows to have a borderline significant explanatory power (B = -.16, p = .068). As this explanatory power can still result in a small mediation effect of passive SME, we will regard this as significant.

(30)

Table 4

Significance of individual IV’s

Independent variables B t p 95% CI

Behavioral SM E-loyalty

Attitudinal SM E-loyalty .127 2.62 .009 .03; .22

Active SME .705 17.04 .000 .62; .79

Passive SME .126 2.35 .019 .02; .23

Transactional behavior

Attitudinal SM E-loyalty .280 3.52 .001 .12; .44

Active SME -.014 -.20 .840 -.15; .12

Passive SME -.161 -1.83 .068 -.33; .00

4.5. Mediation analyses

To test our hypotheses we used the PROCESS macro model number 4 by (Hayes, 2017) to perform a simple mediation model, which involves only one mediating variable. We ran the simple mediation model four times (see Table 5). We want to test how the effect of our independent variable (attitudinal SM E-loyalty (X)) can be apportioned into its indirect effect on our two dependent variables (Y1: Social media behavior, and Y2: transactional behavior) through our mediators (M1: Active SME, and M2: Passive SME) and its direct effect on our independent variables (Y1 and Y2) (Preacher & Hayes, 2008). Lastly, to control for the possible effects of our covariates satisfaction, trust, perceived quality, online shopping experience and social media experience, they were also included into the analyses.

(31)

Table 5

Mediation models: Testing direct and indirect relationships between attitudinal SM E-loyalty, active and passive SME and social media and transactional behavior

Model 1 Active SME (M1) Social media behavior (Y1)

𝛽 SE p 95% CI 𝛽 SE P 95% CI

ASME (M1) - - - - b1 .74 .04 <.001 .66; .81

AEL (X) a1 .26 .09 < .01 .08; .43 c1 .38 .08 <.001 .22; .55 Direct effect c’1 .20 .05 <.001 .09; .30

Indirect effect .19 - - .07; .32

F(6, 211) = 7.18, p < .001; R2 = .17

Model 2 Passive SME (M2) Social media behavior (Y1)

PSME (M2) - - - - b2 .55 .07 <.001 .41; .68

AEL (X) a2 .39 .07 <.001 .25; .54 c2 .38 .08 <.001 .22; .55 Direct effect c’2 .17 .08 <.05 .01; .32

Indirect effect .22 - - .13; .33

F(6, 211) = 7.18, p < .001; R2 = .17

Model 3 Active SME (M1) Transactional behavior (Y2)

ASME (M1) - - - - b3 .00 .06 .99 -.12; .12

AEL (X) a3 .26 .09 <.01 .08; .43 c3 .10 .08 .21 -.06; .26 Direct effect c’3 .10 .08 .22 -.06; .27

Indirect effect .00 - - -.03; .03

F(6, 211) = 4.80, p < .001; R2 = .12

Model 4 Passive SME (M2) Transactional behavior (Y2)

PSME (M2) - - - - b4 -.16 .07 <.05 -.31; -.01

AEL (X) a4 .39 .07 <.001 .25; .54 c4 .10 .08 .21 -.06; .26 Direct effect c’4 .16 .08 .05 .00; .34 Indirect effect -.06 - - -.14; .00 F(6, 211) = 4.80, p < .001; R2 = .12

Notes. SME refers to “social media engagement”; ASME refers to “active social media engagement”; PSME refers to “passive social media engagement”; AEL refers to “attitudinal SM E-loyalty”.

4.5.1. Mediating relationship between attitudinal SM E-loyalty, active SME and social media behavior

Our first simple mediation model tested the relationship between attitudinal SM E-loyalty (AEL) (X), active SME (ASME) (M1) and social media behavior (SMB) (Y1). Our IV, our mediator and our five covariates together explained 17% (R2 = .17) of the variance in our DV social media behavior (SMB), F(6, 211) = 7.18, p < .001, as can be seen in Table 5. Firstly, our results showed that when respondents AEL was stronger, their SME was also stronger as is reflected in (𝛽 c1=.38 p < .001), hereby providing support for hypothesis 1. Secondly, ASME was stronger for respondents who also had a stronger AEL (𝛽 a1=.26, p < .01), hereby supporting hypothesis 2b. Thirdly, our results indicate that ASME has a positive effect on

(32)

SMB as is reflected in (𝛽 b1=.74, p < .001), hereby supporting hypothesis 3B. Finally, in order to establish a mediation effect of ASME on the relationship of AEL and SMB the direct effect should be lower compared to the total effect and that is indeed the case (𝛽 c’1=.20, p < .001).

Furthermore, Hayes (2017) mediation procedure showed that the indirect effect was significant, as the 95% confidence interval did not include zero 95% CI [.07; .32]. Hence hypothesis 4b was supported as well. Including the confounding variables dd not change the interpretation of these results, hereby further validating the findings.

4.5.2. Mediating relationship between attitudinal SM E-loyalty, passive SME and social media behavior

Our second simple mediation model tested the same relationship between attitudinal SM E- loyalty (AEL) (X) and social media behavior (SMB) (Y1), but with another mediator, namely passive SME (PSME) (M2). Our IV, our mediator and our five covariates together explained 17% (R2 = .17) of the variance in our DV social media behavior (SMB), F(6, 211) = 7.18, p <

.001, as can be seen in Table 5. For model 2, same as for model 1, AEL is positively related to SME (𝛽 c2=.38, p < .001). We can conclude that based on the behavioral SM E-loyalty sub- dimension SMB, hypothesis 1 can be confirm. Secondly, AEL had a positive effect on PSME (𝛽 a2=.39, p < .001). Furthermore, the ‘a-path’ of model 2 was not only significant but it was also stronger compared to the a-path of model 1, supporting both hypothesis 2A and 2C.

Thirdly, PSME had a positive effect on SMB (𝛽 b2=.55, p < .001). Furthermore, the ‘b-path’

of model 2 was not only significant but it was also weaker compared to the b-path of model 1, supporting both hypothesis 3A and 3C. Finally, the direct effect was found to be significant (𝛽 c’2=.17, p < .05) and was lower compared to the total effect. Furthermore, the confidence interval of the indirect effect did not include zero 95% CI [.13; .33]. Hence, we can establish a mediation effect with PSME as a mediator, based on which we can confirm H4A. Including the confounding variables did not change the interpretation of these results, hereby further validating the findings.

4.5.3. Mediation relationship between attitudinal SM E-loyalty, active SME and transactional behavior

Our third simple mediation model tested the relationship between attitudinal SM E-loyalty (AEL) (X), active SME (ASME) (M1) and transactional behavior (TB) (Y2). Our IV, our

Referenties

GERELATEERDE DOCUMENTEN

previous model, the effect of temporal orientation to influence stock price increase measuring firms’ performance is not dependent on market competitiveness in this

H1. The upcoming of social networking sites has to an opening of the humanitarian marketing niche to a more interactive and community focused form of marketing. Through this

As mentioned before, the role of the patient has shifted from a passive towards a more active participant within the doctor-patient relationship (Broom, 2005). Based on this, doctors

This study has taken the first steps into discovering the potential of Musical Engagement (ME) in the music industry by exploring its effects on brand loyalty and Social

THE ONLINE TRAVEL MARKET The role of knowledge, privacy concerns, and e-trust to create loyalty in the online travel market: A moderated mediation model.. By Erika

(2001) and Yang and Peterson (2004) state that the perceived switching costs – satisfaction link is an important interaction effect in explaining customer loyalty. ▪ No

Thus, this paper extends previous studies that examined customer loyalty as a one-dimensional construct (for example Zeithaml et al. Results show that the attitude or

Moreover, this research focuses on the Fast Mover Consumer Goods industry (FMCG), the effect of criticism is narrowed by food safety issues and by social media as information