Driving engagement by generating consumer-brand identification : how can service brands benefit from the shift to visual content in social media marketing?

(1)

Author: Mariëlle (M.A.M.) Snoek Student number: 10800956

Date of final submission: 22-07-2018

University of Amsterdam – Amsterdam Business School Supervisor: Gijs Overgoor

Master Thesis 2017/2018 Digital Business

Driving engagement by generating consumer-brand identification

How can service brands benefit from the shift to visual content in social media marketing?

(2)

1

Statement of originality

This document is written by student Mariëlle Snoek (10800956) who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the content.

(3)

2

Figure 1: A conceptual model demonstrating the potential interaction in which the CBI mediates the relationship between brand warmth and user engagement. This mediation is moderated in two ways. Prestige moderates the mediated relationship between brand warmth, CBI and engagement. In addition, it moderates the direct relationship between brand warmth and user engagement. Furthermore, the personal-identity expressiveness motive moderates the mediated relationship between brand warmth, CBI and user

engagement ... 18

Figure 2a: Visual presentation of the Ordinary Least Square (OLS) regression with the conditional Process Model 22. The independent variable is the overall (mean) level of engagement ... 32

Figure 2b: Engagement as a function of Brand prestige and Brand warmth ... 32

Figure 2c: Engagement as a function of the Personal-identity expressiveness motive and the Consumer-brand identification ... 33

Figure 3a: Visual presentation of the Ordinary Least Square (OLS) regression with the conditional Process Model 22. The independent variable is the engagement in terms of likes ... 35

Figure 3b: Visual presentation of the Ordinary Least Square (OLS) regression with the conditional Process Model 22. The independent variable is the engagement in terms of comments ... 35

Tables Table 1: 2 x 2 between-subjects design ... 19

Table 2: Descriptives and correlations between the variables (Cronbach’s Alphas on diagonal) ... 26

Table 3: Regression results on the mediation effect of CBI for all 4 models ... 28

Table 4: Regression results for moderated mediation ... 31

(6)

5

Abstract

As the focus on social media marketing has shifted from presenting textual

information to visual content, service companies with a main focus on offering experiences must embrace and make us of this shift, to overcome the struggle of communicating

intangible benefits. To see whether service brands are able to create consumer-brand identification to drive engagement on social media by posting visual content, I draw upon research regarding the antecedents of this process. Brand warmth is one of the antecedents for consumer-brand identification for brands selling products. I propose that this higher-order, brand-related value can generate the perception of consumer-brand identification for service brands, and consequently result in higher levels of engagement. In addition, I propose that this relationship is moderated by a brand related attribute, namely brand prestige; and a consumer-specific factor, namely the motive to engage with brands on social media. The level of

engagement has been investigated more in-depth, by comparing the effects of the variables on both likes and comments separately. The hypotheses are tested with a sample of 262

participants. Results show that consumer-brand identification partially mediates the relationship between brand warmth and engagement. Furthermore, brand prestige only moderates the relationship between brand warmth and engagement in terms of comments. Additionally, the personal-identity expressiveness motive solely moderates the mediated relationship of consumer-brand identification on brand warmth and engagement in terms of likes. These results add to theory that the process for consumer-brand identification is, at least for the two antecedents studied here, similar for product-selling brands as for service brands. Furthermore, as results for the two type of engagement varies, this provides evidence for the theory that there is an underlying constructs affecting the engagement behavior on social media. Lastly, this study provides insights in the user-specific and brand-specific factors that could have an influence on the process of identification and engagement. For managers of service brands, these results imply that they can benefit from the shift to visual content, as showing brand warmth in pictures is able to increase engagement. Managers must set specific social media marketing objectives in terms of the type of engagement, and when it has

defined its target customer group, it is able to adjust and optimize the social media marketing strategy while keeping in mind their customers motive and their own reputation.

Keywords: brand prestige, brand warmth, consumer-brand identification, engagement, personal-identity expressiveness motive, social media marketing, visual content

(7)

6 1. Introduction

Have you ever seen the following advertisements texts: ‘The dream delivered’, or ‘VU

is looking further’? These are advertising texts used by respectively the airline company KLM

and the Vrije Univeristeit in Amsterdam (KLM; VU). This last slogan even made it among the top four of the qualifications for ‘stupid University advertising slogans’, according to the Platform Hervorming Nederlandse Universiteiten (H.NU). And let’s be honest: how will a potential customer be convinced that the ‘dream will be delivered’, and what does it mean as a University tells you they ‘think further’? Though the slogan ‘think further’ was ranked

second in the qualification back in 2016, the VU has not yet made any alterations. For a long time already, service brands are experiencing difficulties in the communication of their benefits, and they apparently still are. These difficulties are due to the characteristics of a service (Blankson & Kalafatis, 1999). Kotler and Armstrong (2010, p. 248) define a service as “any activity or benefit that one party can offer to another, which is essentially intangible and does not result in the ownership of anything”. Because of the intangibility and the absence of ownership, customers are not able to assess any quality cues or advantages in advance (Lovelock & Gummesson, 2004). In order for service brands to differentiate from their competitors, and to charge a premium for their offerings, they must be able to build a strong brand (Biedenbach & Marell, 2010).

Building a strong brand can be hard, but nowadays, social media provides the opportunity to businesses to get closer to their customers than ever. These platforms allows businesses to strengthen the relationship with customers and to receive immediate feedback from them (Virtanen, Björk, & Sjöström, 2017). Presently, with the advancements in the visual technology, a shift from textual to visual content is taking place (Zhou, 2017).

Businesses start to lay greater emphasis on the use of visual-based social media platforms in order to express their brand image (Virtanen et al., 2017). The concept that is considered as a key feature on these platforms is the engagement, which demonstrates the positive aspect of the interaction (Attfield, Kazai, Lalmas, & Piwowarski, 2011).

As engagement is regarded as an important factor for raising brand awareness, research has analyzed which factors could drive this engagement (Hutter, Hautz, Dennhardt, & Füller, 2013). Consumer-brand identification has been found as an important driver of consumer engagement on social media platforms (Jamal & Goode, 2001). As research has investigated what the antecedents are of this process of consumer-brand identification, this

(8)

7

research was limited to the investigation of brands selling products (Stokburger-Sauer, Ratneshwar, & Sen, 2012). Due to the intangibility of a service, it appears that there is a distinct difference in offerings between brands selling products, and brands with experiences as their core offering. Therefore, research is needed to analyze whether the antecedents of consumer-brand identification are the same for service brands (Lovelock & Gummesson, 2004). Hence, this study will investigate two of the six antecedents found by Stokburger-Sauer et al. (2012).

As the focus of this study is on services with experiences as their core offering, the variable brand warmth will be analyzed as an antecedent of consumer-brand identification, as this is an example of an emotional advertising appeal depending upon the experiential part of consumption (Albers-Miller & Royne Stafford, 1999). The second antecedent that will be investigated is the brand prestige. Although this variable showed the smallest predicting power for brands offering products, the prestige of a brand is considered as an important cue for customers in order to make an attempt in assessing the quality of a service brand upfront (Baek, Kim, & Yu, 2010).

Although previous research has demonstrated a positive relationship between consumer-brand identification and engagement, so far, research has not taken into account any user-specific factors that could be influencing this relationship. The user- and gratification theory is used to explain the differences in motives individuals have to use social media (Muntinga, D. G., Moorman, M., & Smit, 1973). More specifically, social media users even appear to have different motives to engage with brands online (Shao, 2009). As not all individuals engage with brands in order to express their identity, this motive could be influencing the effect of consumer-brand identification on the level of engagement.

Moreover, as the concept of engagement entails both the liking and commenting behavior of social media users, research has neglected the view that there are underlying constructs that influence the type of engagement in which the individuals will participate (Kabadayi & Price, 2014). Therefore, this study will investigate whether the consumer-brand identification and its proposed antecedents have any differences in effect on the two distinct types of engagement.

The goal of this study is to find out whether service brands could benefit from the shift from textual to visual content in order to create consumer-brand identification and

subsequently to increase the engagement. This goal will be accomplished by answering the following question: How does consumer-brand identification mediate the relation between

(9)

user-8 specific factor, such as the personal-identity expressiveness motive, influence this

relationship? In addition, there will be analyzed whether these factors have different effects,

depending on the type of engagement.

By answering this question, this research contributes to literature by examining whether the process of consumer-brand identification is comparable among retail brands and service brands. Furthermore, it does not only investigate the effectiveness of the different visual content appeals on engagement, but it also takes into account brand-specific, and user-specific attributes. Moreover, as the effect of the constructs will be compared among the two types of engagement, theory will gain insights whether there is indeed an underlying construct

influencing the motives to participate in a certain type of engagement. Managers could use these results to review, adjust and optimize their social media marketing strategy.

In order to test the hypotheses, participants will be a sample of Instagram users, as this is considered as the platform that is particularly designed for visual content (Chen, 2017). Data will be collected by means of an experimental survey design. Because this research pays specific attention to consumer-brand identification of service brands with experiences as their core offering, participants will randomly be shown one of two different posts from an airline company, having either low or high prestige. Ordinary Least Square regression analyses will be run to test the proposed relationships.

The remaining of this paper is divided in different sections. The literature review will introduce the concept of social media marketing; explain what is known about the drivers of engagement on the visual-based platforms and how these could differ for product- and service-related brands. Furthermore, the theoretical framework will outline how the main variables will interact, and hypotheses are constructed to introduce the expected relationships. Subsequently, the method section will describe the design, data collection, measurements and statistical procedure in order to test the hypotheses. This section will be followed by the results and a discussion of the findings. Finally, the contributions of the study and limitations will be given together with the recommendations for further research. The conclusion will link the findings back to the research question.

2. Literature review

In this section, the concept of social media marketing will be introduced as will be the shift from textual to visual content. In addition, a short explanation about the engagement

(10)

9

on these social media platforms will be provided. Furthermore, there will be explained how the process consumer-brand identification can lead to engagement and how this identification process may differ among brands selling products, and brands offering services with their focus on the experiences. Lastly, some factors that could be influencing the engagement on social media platforms will be introduced.

2.1 Social media marketing

With the proliferation of the Web 2.0 technologies, social media provide a variety of applications and platforms for individuals to enable interactions, collaborations and the creation and distribution of content (Farook & Abeysekara, 2016; Kim & Ko, 2012). With the exponential increase in the use of these platforms, businesses have started to use social media as new advertising and marketing tools for purposes such as promoting products and services, or increasing the relationships with customers (Farook & Abeysekara, 2016). The social media marketing concept delineates the acts of using the social platforms for marketing activities by creating a connection between brands and consumers, while offering them a personal channel to network and attain social interaction (Chi, 2011). Businesses have begun to recognize the power of the social networks as the internet is perceived as the enabler of communication, by being an open, cost-effective and omnipresent network (Farook & Abeysekara, 2016). With its capabilities of interactivity, its wide reach, flexibility and rapid processing ability, it circumvents the geographical barriers and removes the physical distance between organizations and consumers (Farook & Abeysekara, 2016). This allows businesses to strengthen their relationship with customer by engaging in real-time, interactive

communications with them (Kim & Ko, 2012). This real-time communication not only generates feelings of trust and intimacy among customers, but the immediate feedback could also improve businesses products, services and business models.

Currently, social media marketing is associated to relationship marketing where a shift is taking place from ‘trying to sell’ to ‘building relationships’ with consumers (Farook & Abeysekara, 2016). Hence, it is not that much about the ‘big campaigns’ anymore. It are rather the ‘small acts’ of marketing activities that may change the attitude of customers, and can ultimately lead to the achievement of the business’ marketing objectives (Farook & Abeysekara, 2016; Virtanen et al., 2017).

So far, research shows that social media marketing activities can significantly influence customer’s attitude, and ultimately impact the financial and reputational performance of a firm (Kim & Ko, 2012; Kumar, Bezawada, Rishika, Janakiraman, &

(11)

10

Kannan, 2016; Virtanen et al., 2017). It plays a key creating brand awareness and helps customers in acquiring information, forming opinions and shaping their attitudes, purchase behavior and post-purchase evaluation and communication (Mangold & Faulds, 2009).

However, social media platforms can take several forms, such as weblogs, microblogs, social blogs, platforms for video’s and pictures, podcasts and many more (Kim & Ko, 2012). Within these platforms, a distinction can be made between relational-based platforms, and visual based platforms (Zhou, 2017). Today, businesses start to lay greater emphasis on the use of visual-based social media platforms. With the current advancements in the visual technology (for instance the high pixel density and the 5K display) and the conveniences in visual communication (such as the extremely fast information-processors and easy image editors), visual content have become more important in terms of branding (Zhou, 2017). In addition, consumers are not only interested in the functions, quality and price of a product or service anymore. The emphasis of advertising has changed towards showing the story, character and lifestyle behind a company and visual-based social media platforms such as Instagram and Pinterest are specifically designed to express unique brand images and share visual brand stories (Chen, 2017). Concerning advertising effectiveness, pictures have a well-known superiority over words when it comes to learning in terms of recognition responses, recall responses and memory; and they are more effective in persuading positive brand evaluations in contrast to verbally oriented advertisements (Rossiter & Percy, 1978). Besides, imagery is perceived as one of the prominent ways for communicating online (Bakhshi, Shamma & Gilbert, 2014), therefore the focus of this study will be on visual content.

2.2 Social media engagement

Regarding the activities on social media, the objective that is considered as the most important is the engagement of customers (Hollebeek, 2011; Hollebeek, Glynn & Brodie, 2014; Van Doorn Lemon, Mittal, Nass, Pick, Pirner, & Verhoef, 2010). It is argued that increasing levels of consumers’ engagement lead to increased brand awareness, word-of-mouth activities and purchase intention (Hutter et al., 2013) This can ultimately be beneficial for the realization of favorable organizational performance outcomes, such as sales growth, brand referrals, enhanced co-creation with customers and superior profitability (Hollebeek et al., 2014).

User engagement is a key concept in social media marketing referring to ‘the quality of the user experience that emphasizes the positive aspect of the interaction’ (Attfield et al., 2011, p.9). Engagement is mostly assessed in the number of likes and comments. Liking is a

(12)

11

strong social signal that demonstrates the extent to which the photo appeals to the user, whereas commenting is considered as an explicit action on the content (Bakhshi et al., 2014). According to previous literature, this type of behavioral response is a form of engagement, as it stipulates the amount of effort someone puts in interacting with the brand (Hollebeek, 2011; Hollebeek et al., 2014; Van Doorn et al., 2010).

However, what most research neglects nowadays is that the drivers for commenting and liking on social media can be different. According to Kabadayi, and Price (2014), it depends on someone’s mode of interaction whether he/she will like or comment on a brand’s post. Commenting on a brand’s post includes giving an opinion about the brand. In addition, someone’s name and profile picture will be visible to others. This is in contrast to liking a brand’s post where the visibility is less, as it will only show a username, instead of the user’s picture and opinion (Kabadayi & Price, 2014). This is why this study will investigate the effect on overall engagement, as well as the liking and commenting behavior separately.

There has already been some research on the drivers of engagement on Instagram. For example, Chang, Yu and Lu (2014) argued that the persuasiveness of a post leads users to engage with the message. They found that argument quality, post popularity and

attractiveness all have an influence on the user’s engagement (Chang et al., 2014). Furthermore, Geurin and Burch (2017) stated that posts with a focus on the brand’s uniqueness rather than its low-cost offerings, resulted in more engagement.

A qualitative research by Chen (2017) studied young consumer’s perceptions of social media marketing. This study showed that refined content, privileged information, newness and social or celebrity endorsement were all factors positively influencing young consumer’s attitudes towards posts. The author recommends using marketing communications that makes consumers feel good about themselves, by using images that enhance consumers’ self-image and social interaction experiences.

2.3 Consumer-brand identification

A way for companies to enhance consumer’s self-image and to create social interaction experiences is by ‘visual branding’. Instagram has given rise to this new

phenomenon that is defined as the act of using a variety of visual elements to strengthen and convey the brand image (Eriksson, Andersson, & Karlsson, 2015). On Instagram, companies not only present textual information, but also try to show users the identity and image of the brand by telling their story visually (Virtanen et al., 2017). When encountering this content, the consumer can either identify with the image conveyed by the brand, or they do not, based

(13)

12

upon their level of consumer-brand identification (Johar & Sirgy, 1991). As brand and possessions help in identifying someone’s uniqueness, they are considered as influential factors in creating, and influencing a consumer’s identity (Tuškej, Golob & Podnar, 2013). By identifying with the brand, an individual differentiates itself from users not identifying with the brand.

Research by Johar and Sigry (1991) shows that the level of consumer-brand identification could have an impact on the effectiveness of different marketing appeals. Brands expressing their individuality can function as tools for social integration and are able to symbolize a user’s personal accomplishments (Ruane & Wallace, 2015). Sook Kwon, Kim Sung and Yun Yoo (2014) argue that when users come across a brand’s message, they will determine whether they can link their own self-concept with the imagery shown by the brand. If so, the user may hold positive attitudes towards the brand. Kim, Lee and Ulgado (2005) also support this statement. They found that if a consumer can match their self-concept with that of a brand, it sparks emotions such as affection, dignity and pleasure, ultimately creating a long-term consumer-brand relationship through brand attachment. Jamal and Goode (2001) add to this that brand identification is valued as a predictor of consumer’s brand preferences and consumer’s satisfaction.

As consumer-brand identification is a decent predictor of a user’s brand preferences, it also might lead to engagement on social media platforms. Akerlof and Kranton (2000, p. 717) claim that “because identity is fundamental to behavior, choice of identity may be the most important economic decision people make”. This is in line with research that states that consumer-brand identification is could lead to consumer loyalty and advocacy (Stokenberg-Sauer et al., 2012). Research on social media platforms showed that consumer-brand identification was one of the greatest predictor for increasing the likelihood of engagement (Kim, Sung, & Kang, 2014). Users will participate in a brand’s conversations and activities once they feel their personality is similar to the brand (Bagozzi & Dholakia, 2012).

As consumer-brand identification could lead to pro-brand consequences, such as consumer loyalty, advocacy and engagement, Stokburger-Sauer et al. (2012) took it a step further by investigating the aspects that could drive the consumer’s identification with a brand. They defined consumer-brand identification as “the consumer’s perceived state of oneness with a brand” (2012, p.407). They found six factors that could drive the consumer-brand identification of customers. The first three (mainly cognitive-based) are consumer-brand-self similarity, brand prestige and distinctiveness, and the other three (more affective in nature) are brand social benefits, brand warmth and memorable brand experiences.

(14)

13

However, most research on consumer-brand identification has been restricted to brands offering products, and Stokburger-Sauer et al. (2012) justifiably mention that the process for brand-identification may differ for a brand offering services with a focus on the experience of the service. Products are an easy way for individuals to show who they are, or who they want to be by using these products as symbols (Arbore, Soscia & Bagozzi, 2014). Literally, this concept is called the extended self and is defined as “the self created by external objects with which one surrounds oneself” (Solomon, 1994, p. 620). Individuals identify more easily with brands selling products because the products help them to differentiate, and promote the individual’s sense of identity by forming their extended selves (Arbore et al., 2014).

Conversely, for services with their main focus on offering experiences, there is usually no product present to leverage the extended self (Arbore et al., 2014). Services differ from products on a bi-dimensional concept of intangibility. This concept contains a physical

dimension, relating to the materiality of the product or service; and a mental dimension that is concerned with the difficulty in defining, formulating or understanding the product or service in a sufficient way (Lovelock & Gummesson, 2004). The intangibility leads to the fact that services cannot be seen, heard, tasted, smelled, or felt before purchase, and therefore are more difficult to evaluate before consumption (Lovelock & Gummesson, 2004). Positioning for service brands is harder because they need to communicate vague and intangible benefits (Blankson & Kalafatis, 1999). These difficulties in evaluating a service may make it harder for individuals to judge whether they identify with a brand or not.

As a result, contradictions exist between what kind of advertising appeal works best for service companies (Utkarsh, Maurya, & Medhavi, 2016). The appeals used in an

advertising strategy are an intrinsic part of the marketing approach in order to influence consumer behavior and the process of consumer-brand identification (Johar & Sirgy, 1991). Emotional advertising appeals are based upon the emotional and experiential part of

consumption and seek to create either positive or negative feelings about a brand to foster customers into desired behaviors (Albers-Miller & Royne Stafford, 1999). Showing brand warmth is perceived as an example of an emotional advertising appeal. In contrast to the rational appeal that tries to convince the public by emphasizing a product’s benefits, showing brand warmth entails the creation of a likeable or friendly brand (Albers-Miller & Royne Stafford, 1999).

Besides the intrinsic part of the marketing approach, there are other factors that could guide consumer’s behavior and their brand identification. For example, it is proposed that brand-related factors such as the trustworthiness, uniqueness, distinctiveness and prestige of a

(15)

14

brand could also influence the extent to which someone identifies with a brand’s values (Bhattacharya & Sen, 2003; Stokburger-Sauer et al., 2012). However, results regarding the effect of a brand’s prestige in the product-centered businesses are mixed. Stokburger-Sauer et al. (2012) indicate that the effect of prestige could be higher in conventional status categories. As for the service sector, the consumer’s choice is often based upon the level of prestige in order to get an indication of the quality, the effect of this variable will be investigated more deeply in this study (Baek et al., 2010).

In addition, another factor that could influence the effectiveness of marketing activities that is outside the control of marketing managers are the different motives individuals could have to use social media (Muntinga, et al., 1973). More specifically, the motives to follow and interact with brands online could even differ between individuals (Shao, 2009). In the next section, these differences in motives and their effect will be explained more thoroughly.

3. Theoretical framework

In this part, the possible relationships between the key variables will be outlined. First, there will be explained how brand warmth could generate consumer-brand identification. Secondly, how the brand-related attribute prestige could strengthen the direct relationship of brand warmth on engagement, as well as the indirect relationship through consumer-brand identification, will be explained. At last, the user-specific factor, namely the personal-identity expressiveness motive will be taken into account and the possible relationships between all these concepts will be explained in detail followed by the related hypotheses. Figure 1 shows the visual presentation of the proposed relationships.

3.1 Brand warmth and consumer-brand identification

As mentioned, emotional advertising appeals are based upon the emotional and experiential part of consumption and brand warmth is considered as one example of this kind of appeal (Albers-Miller & Royne Stafford, 1999). Brand warmth is conceptualized as a brand having positive intentions towards its customers (Bernritter, Verlegh, & Smit, 2016). As researched by Stokburger-Sauer et al. (2012) brand warmth is one of the antecedents of customer-brand identification in the retail sector. They argue that the creation and

maintenance of an identity are intrinsically affective processes where individual’s emotions play an essential role. This is in line with research from Tripp (1997) who found that

(16)

15

the advertisements. In addition, he argued that for brands offering services, rational appeals work best.

Conversely, on Instagram, the main goal of the posts is to visually story-tell the message and to express the brand’s identity (Virtanen et al., 2017). Individuals tend to distinguish brand messages on the same spectrum as they do with people: warmth and competence (Bernritter et al., 2016). As mentioned, services are hard to judge on its quality and the personnel’s competence before consumption and therefore, the identification process will depend more strongly on the brand warmth (Lovelock & Gummesson, 2004). In addition, O'Cass and Grace (2004) mention that with services, instead of abstract images, customers are more able to identify with brand imagery presenting the people-centered nature of the service. They argue that marketing managers should focus more on showing the ‘friendly face’, instead of images impairing the feelings of personal interaction. This statement is also

supported by Stokburger-Sauer et al. (2012) who claim that it is easier to identify with higher-order, identity related brand meanings such as pictures showing brand warmth, than with lower-order meanings that come along with conveying concrete brand features. Furthermore, research found that service advertisements often contain more emotional appeals than do product advertisements, to improve the tangibility of the service offered, which could help to facilitate the identification process (Blankson & Kalafatis, 1999).

Therefore, as opposed to posting concrete brand features, it is proposed that when service companies reflect the warmth of their brands in posts, they will leverage identity-related brand meanings where a customer can identify with. As a consequence, once a consumer identifies with the brand, they will be more likely to engage with the brand by liking or commenting. Consequently, the next hypothesis follows:

Hypothesis 1: consumer-brand identification mediates the relationship between brand warmth and user engagement

3.2 Brand prestige as a moderator

Though the results regarding the effect of a brand’s prestige on consumer-brand identification in the product-centered businesses are mixed, Baek et al. (2010) consider the prestige of a brand as an important quality cue for customers in order to assess the value of service brands upfront. Prestige can be defined as a brand’s perceived exclusivity and status, and is often conceptualized as the perception that other stakeholders, whose opinions are respected, consider the brand as valuable and respected (Kuenzel & Vaux Halliday, 2008).

(17)

16

However, as Bernritter et al. (2016) explained, individuals tend to place brands on the same spectrum as they do with people: warmth and competence. Furthermore, they argue that the first assessment will be based upon the warmth. As for brands high in prestige, this first assessment will even be more about the brand warmth, as the level of prestige already gives potential customers an indication of the value offered, as quality appears to be directly linked with prestige (Baek et al., 2010). As the first assessment will be positive, such that the brand shows its loving character, the potential customer can show his/her affinity by engaging with the post. Hence, the following hypothesis follows:

Hypothesis 2: the prestige of a brand moderates the relationship between brand warmth and engagement, such that high levels of brand prestige will increase the effect of brand warmth in a post on the likelihood of engagement

In addition, Baek et al. (2010) mention that brand prestige is often associated with an individual’s self-concept. The need for identification with a brand can either come from an individual’s intrinsic or extrinsic motives (Smidts, Pruyn, & Van Riel, 2001). According to the self-concept research, and in light of the self-enhancement theory, people tend to perceive themselves in a positive light. Consumers will seek to exhibit in behaviors that involve the maintenance and confirmation of positive self-views, which indicates the intrinsic motive of identification (Bhattacharya & Sen, 2003). As a brand conveys brand warmth in its posts, it shows the brand is very loveable. Individuals may perceive their own identity as similar to the brand, and identify with it. The prestige of a brand could reinforce the relationship between brand warmth and the identification process of an individual. If the brand is valued as lovable by other stakeholder, individuals may feel the desire to associate with this brand to increase their self-concept (Kuenzel & Vaux Halliday, 2008). Stokburger-Sauer et al. (2012) argue in their research that it should not be surprising that users try to achieve the identity-related need, by identifying with prestigious social entities, such as brands.

A second research stream that leverages support for this relationship is the social identity theory. A basic extrinsic motive for identification that is derived from this theory is the individual’s need for self-categorization (Smidts et al., 2001). The process of associating with a certain group, entity or brand may feel rewarding for the individual, as it reinforces a desired social identity by demonstrating their membership in a particular social category. When a brand is able to create warm feelings among a certain community with its social media post, it could support an individual in identifying with the brand and defining one’s

(18)

17

place in society (Carlson, Todd Donavan, & Cumiskey, 2009). As the brand is regarded as having prestige, the individual may want to categorize his/herself within this group and emphasize this towards others. As users want to express their identity, they will be more likely to engage with messages of brands where they identify with, as this will expose their username on the social media platform to other users (Kabadayi & Price, 2014).

Consequently, the prestige will strengthen the relationship between brand warmth, consumer-brand identification and engagement. Accordingly, the following hypothesis is proposed:

Hypothesis 3: the prestige of the brand moderates the mediated relationship between brand warmth, consumer brand identification and engagement, such that brands with more prestige will enhance the effect of brand warmth on consumer-brand identification and ultimately leads to engagement

3.3 Personal identity expressiveness motive as a moderator

Whether consumer-brand identification leads to engagement on social media, could be dependent upon the specific motivation of the consumer to engage in online brand activities. The user and gratification theory can be used to investigate the customer-centric perspective on social media (Muntinga et al., 1973). This approach is used to discover consumer’s motives for using certain media and the consequences that come along with these motives (Katz, Blumler, Gurevitch, 1973).

Motivations for social media usage differ among the kind of consumer’s online activities that can be categorized along three dimensions: consuming, contributing to and creating brand-related content (Shao, 2009). Consuming online content represents the minimum level of online activity in relationship with brands. It encompasses only the

consumption of brand-related content, such as videos, product reviews or dialogues on brand forums (Muntinga et al., 1973). Contributing to brand related content means that people may participate in the conversation through interacting with the content (Shao, 2009). Creating brand-related content represents the ultimate level of online brand-related activity. Users who create brand-related content write weblogs and/or product reviews, upload videos about the brands products or services or write articles about it. The focus of this study is on

engagement, and thus on the ‘contributing’ dimensions of consumer’s online brand related activities.

(19)

18

Muntinga et al. (2012) found in their research that the contributing type of consumer’s online brand activities can be driven by three distinct motivations: personal identity, integration and social interaction, and entertainment. If a customer does identify with the post shown by a company, but their main motivation to engage with a company is merely entertainment driven, the customer may not engage with the post by liking or commenting on it. In

comparison, if a follower identifies with the company, and his/her main motivation to engage with brand online is to express or shape one’s identity, the user will be more likely to like or comment on the post. Therefore, the following hypothesis is proposed:

Hypothesis 4: personal-identity expressiveness motive moderates the relationship between consumer-brand identification and user engagement, such that users with a high personal identity motive will have higher levels of engagement once they identify with a brand than users with a low personal identity motive

(20)

19 4. Method

4.1 Research design and sample

Participants were conducted using convenience sampling. For the Netherlands, it is reported that approximately 11.5 million people, from the age of 15 and older, are using social media. Instagram accounts for 4.1 million users, and their main target group is between 15 and 39 years (Nationaal Social Media Onderzoek NewCom, 2018). Therefore, the participants targeted were Dutch consumers who are active on social media and above the age of 15. An Internet distribution method was chosen to reach an as large sample as possible. A total of 269 complete surveys were collected.

The proposed hypotheses were tested in an online experimental survey using a 2 (no brand warmth versus brand warmth) x 2 (low prestige versus high prestige) full factorial between-subjects design, illustrated in Table 1. This implicates that there were in total four treatments for the experiment. For each of them, a survey is designed which can be found in the Appendix. Participants were randomly assigned to one of the four treatments, which is important to eliminate other sources of systematic variation so that any variation in the research will be due to the manipulation of the independent variables (Field, 2009). In total, two independent variables were manipulated.

Table 1: 2 x 2 between-subjects design

No brand warmth Brand warmth

Low prestige Treatment 1 Treatment 3

High prestige Treatment 2 Treatment 4

4.2 Stimulus development

The focus of this study is on social media marketing with a particular emphasis on visual content. Therefore, the hypothesis will be tested by showing the participants four different types of Instagram posts. Instagram is an application that allows users to edit and share pictures and videos from their mobile devices (Virtanen et al., 2017). As of 2017, Instagram has over 800 million monthly active users, sharing together an average of 95 million photos and videos per day (Mathison, 2016). Instagram is also a popular platform for businesses; for 2017, there are 25 million business profiles on Instagram, and 80% of the users follow at least one business. Instagram has the potential to be used for marketing purposes as 60% of users indicate that they discover new products by using Instagram, and

(21)

20

75% takes immediate action (for example, visiting a company’s website) after seeing a brand’s post (Mathison, 2016). Therefore, this medium was considered as a valid medium to test whether brand posts with different visual content have a different effect on user

identification and engagement.

In addition, because this research pays particular attention to consumer-brand identification of service brands with their main focus on offering experiences instead of products, the posts that were shown to participants were posts from an airline company. The services airline companies offer, are perceived as intangible-dominant, and cannot be

physically possessed. Airline travels are hard to evaluate upfront, and their benefits can only be experienced and assessed afterwards (Shostack, 1977). Lastly, because the airline industry is a highly competitive sector, the power of branding in this industry is well documented, which makes it an interesting industry to investigate (So, King, Hudson, & Meng, 2017)

In order to rule out any differences in results due to brand equity of real brands, a fictive airline company was created along with a fictional name (Flyaway) and logo. A pre-test showed that the name and logo were perceived as credible. In addition, the photo was added ‘a few seconds ago’, so any number of likes/comments that the post could have received would not influence results. Furthermore, with initial interactions, customers tend to align different colors with specific traits. 62-90 Percent of customer’s assessment of, and attitude towards a brand are based on colors alone. Therefore, the color grey was used in the logo in order to eliminate any differences in results based on color (Singh, 2006). Grey is perceived as the most neutral color, hence, participants’ opinion about brand personality, quality, price or service levels would not be influenced through color associations.

4.3 Manipulation design

4.3.1 Independent variable brand warmth. Brand warmth was measured on a nominal level. There were two levels: brand warmth or no brand warmth. Participants in the pre-test were asked to assess six pictures of airline companies on the variables that are discovered as antecedents of consumer-brand identification for product companies, namely: unique identity, brand distinctiveness, brand prestige, brand-self similarity, brand warmth and social benefits. These variables were measured with three items on a 5-point Likert scale. The picture that scored highest on the variable brand warmth and most neutral on the other variables was chosen as the picture showing brand warmth. The picture that scored lowest on brand warmth and most neutral on all other variables was chosen for the ‘no brand warmth’ treatment. Brand warmth was measured on a 3-item, 5-point scale ranging from ‘totally disagree’ to ‘totally

(22)

21 agree’, and the reliability of this scale is .85 (Stokburger-Sauer et al., 2012). An example item

is ‘This Instagram post creates warm feelings among its users’.

4.3.2 Independent variable brand prestige. Brand prestige was manipulated by means of a vignette. A vignette is “a short, carefully constructed description of a person, object, company or situation” (Atzmüller & Steiner, 2010, p.128). It presents a hypothetical situation with a systematic combination of characteristics to which participants respond by revealing their impressions of events, social norms, or values. A pre-test was done in order to test if the vignette was appropriately designed to manipulate the participants. For each treatment, three different vignettes were created and 15 participants were asked to indicate on a 3-item scale to indicate whether the perceived the company described as prestigious or not. The 5-point scale ranged from ‘totally disagree’ to ‘totally agree’. An example item is: ‘This company is a

first-class, high-quality brand’. The reliability of this scale was .92 (Stokburger-Sauer et al., 2012).

4.3.3. Manipulation checks. To see if the manipulation in this experiment worked successfully and to see if the participants got the manipulation as intended, two questions were added. The first question checked whether the participants perceived the brands as having either low or high prestige. Participants were asked to answer the question ‘This photo

was posted by a company with:’ with either ‘low prestige’ or ‘high prestige’.

In order to investigate whether the manipulation for the brand warmth was perceived as intended, a second question was added. Participants were asked to indicate whether the photo that was posted by Flyaway could be considered as either ‘emotional’ or ‘rational’.

4.4 Measurements

4.4.1 Dependent variable user engagement. User engagement was measured on a five-point scale ranging from ‘extremely unlikely’ to ‘extremely likely’. The two items that were measured are ‘How likely are you to click LIKE for the post?’ and ‘How likely are you to COMMENT on the post?’. These two questions measured the features that represent the most fundamental aspects of engagement on Instagram (Bakhshi et al., 2014). For the analyses, the mean of these two variables was computed to analyze the effects on overall engagement. The reliability of this scale was .697. However, because the motivations for liking and

commenting on a post differ, the effects on the separate constructs was also be taken into account.

4.4.2 Mediator variable consumer-brand identification. The consumer-brand identification was measured with a 7-point Likert scale ranging from ‘totally disagree’ to ‘totally agree’. The scale was originally created by Podnar, but the adapted version by Tuškej

(23)

22

et al. (2013) was used. An example item is ‘I feel that my personality and the personality of this brand are very similar’. The scale consists of three items and the reliability was .895.

4.4.3 Moderator variable personal identity expressiveness motive. The variable personal-identity expressiveness motive was measured on a seven-point Likert scale ranging from ‘totally disagree’ to ‘totally agree’. The scale consisted of three items and was initially developed by Thorbjørnsen, Pedersen and Nysveen (2007). However, a slightly different version was used in this study to fit in with the context. An example item is: ‘I engage with companies on social media to express my personal values’. The reliability of the scale was .897.

4.4.4 Control variables. Service category involvement and frequency of Instagram usage were controlled in order to ensure that the results are not caused by any side-effects.

Instagram usage was measured by asking participants how often they use Instagram on a daily basis on a scale ranging from 1 to 5, never to very frequently. Service category involvement was measured on four items on a 5-point Likert scale ranging from ‘totally disagree’ to ‘totally agree’. An example item is: ‘I am very interested in anything related to travelling

services’. The reliability of the scale was .84 (Stokburger-Sauer et al., 2012). Furthermore,

demographic variables were included in the survey to assess the similarity among the sample. These variables included age, gender (male/female), and the highest completed educational level (primary school, high school, MBO, HBO and University).

4.5 Procedure

Questionnaire responses were collected in an indirect manner, approaching potential participants by means of an Internet Distribution method. This approach was chosen over a direct method to reach an as large and heterogeneous sample as possible. The survey was only distributed in English to overcome any bias in the translation to Dutch. Dutch consumers who are active on social media were asked to take part in a survey about consumer behavior on social media. Participation was voluntary for all users, and confidentiality was assured. Data was collected over a period of two weeks. Participants were randomly allocated to only one of the four treatments: an Instagram post of a high prestige brand showing no brand warmth; an Instagram post of a high prestige brand showing brand warmth; an Instagram post of a low prestige brand showing no brand warmth; or an Instagram post of a low prestige brand showing brand warmth. When the social media users agreed to participate, the survey took about 4 minutes to fill out.

(24)

23

4.6 Analyses and predictions

To pre-test whether the manipulations will work out as they should, a pre-test among 15 respondents will be executed. After this pre-test, data will be collected and a data analysis will be conducted with SPSS to test the main hypotheses. First, to see if the manipulations work successfully, a chi square test of independence will be calculated. After this, there will be investigated with an independent samples t-test whether the age is comparable among the four different treatment groups. This randomization will also be analyzed with regards to the participant’s gender and their highest level of completed education with a Chi-square test. Once there is concluded that the variables are all normally distributed, the reliability of the scales will be analyzed by computing the Cronbach’s alpha. To investigate the main

interactions between the variables, the correlations will be computed and an overview of the descriptive statistics will be provided to create an overview of the sample and the results.

In order to test the hypothesis on mediation, four linear regression analysis will be conducted. More specifically, there will be analyzed whether brand warmth predicts consumer-brand identification and overall engagement. In addition, it will be analyzed whether consumer-brand identification predicts user engagement and whether the combination of these two independent variables also have an effect on the overall

engagement. I expect there to be a positive effect between all the variables, in order to argue for a mediation effect of consumer-brand identification on the relationship between brand warmth and engagement.

In order to analyze whether this mediation is moderated by either brand prestige or the personal-identity expressiveness motive of users, ordinary least squares (OLS) regression with the conditional process modeling (PROCESS) program for SPSS will be used (Hayes, 2017). This model uses a bootstrapping procedure to generate a sample size of 5.000 in order to assess regression models and 95 percent bias corrected intervals to investigate any (in)direct effects (Lessard-Bonaventure & Chebat, 2015). Specifically, Model 22 will be used to test the interaction between X (brand warmth) and W (prestige) on M (consumer-brand identification) and the three-way interaction between X, W and M in affecting Y (engagement). In addition, this model also tests the interaction effect between M (consumer-brand identification) and Z (personal-identity expressiveness motive) on Y (engagement). The conceptual and statistical model of Model 22 can be found in the Appendix. Prior to analysis, all numerical variables will be mean-centered to reduce problems associated with multi-collinearity in moderated regression (Tabachnick & Fidell, 2013). Model 22 will be conducted three times to investigate whether there are any differences in results between the two types of engagement, and the

(25)

24

overall average level of engagement, whilst controlling for service category involvement and Instagram usage. I expect there to be a positive interaction effect of brand warmth and brand prestige on both consumer-brand identification and the level of engagement. Furthermore, I expect that the personal-identity expressiveness motive positively interact with the consumer-brand identification such that the higher this motive, the greater the effect of consumer-consumer-brand identification on the different levels of engagement.

5. Results

5.1 Sample

Data was collected of 269 participants. Among these, 113 are male (42.01%) and 156 are female (57,99%). The age of the participants ranged from 17 to 67, however the majority of the participants are aged between 21 and 30 (81%). In addition, the majority of the

respondents indicated that university was their highest level of completed education (45.5%), followed by university of professional education (27.9%). Subsequently, when the

participants were asked about their Instagram usage on a daily basis, most of them indicated that whenever they have the chance, they always check their Instagram (25.3%).

5.2 Manipulation checks

To see if the manipulation in this experiment worked successfully and the participants got the manipulation as intended, a Chi Square test of independence was calculated.

Participants were asked to indicate whether they perceived the brand Flyaway as having high or low prestige. A significant interaction was found between this measure and the variable actually indicating whether the participants were presented the high/low prestige text (²(1) = 190.565, p < .01). This indicates that the level of prestige was manipulated as intended. The same holds for the level of brand warmth that was presented in the picture. The variable that measured the participants perception of the picture as more emotional or rational had a significant interaction with the photo that was actually shown to the participant (²(1) = 216.757, p < .01). Hence, participants who were shown the picture with the airplane regarded the post as showing no brand warmth, the picture with the little child and flight attendant was regarded as showing brand warmth.

5.3 Normality and Randomization checks

All the variables were checked for normality based on their skewness values. The skewness levels of all variables were between minus one and plus one. Hence, the variables used in this study can be considered as normally distributed. Furthermore, to test for

(26)

25

multivariate normality, Mahalanobis Distance was computed for each case in the dataset. The mean of the squared Mahalnobis distance, also known as Mardia’s coefficient was 2.183, which is smaller than 35 (p(p+2), where p = number of variables). Therefore, the multivariate distribution of the variables can also be considered as normal.

To see if the age of the participants was comparable among the four different

treatments, an independent samples t-test was executed. The results show that the mean age of the participants in the condition with the brand having low prestige (M = 25.70, SD = 8.28) was not significantly different from the age of the participants who saw the Instagram posts of the high prestige brand (M = 25.70, SD = 8.15), t (263) = -.004, p = .997. In addition, the age of the participants who saw the picture including brand warmth (M = 25.27, SD = 6.78) also did not differ significantly from the age of the participants who were exposed to the Instagram picture showing no brand warmth (M = 26.09, SD = 9.34), t (263) = -.004, p = .997.

To check whether there were any differences in gender and education among the four treatment groups, a Chi-square test was conducted. For gender, the results indicate that there was no significant difference in gender between the two different groups regarding brand warmth ²(1) = 3.344, p = 0.067. In addition, there was no significant difference in gender between the low and high prestige groups ²(1) = 4.28, p = 0.513. These results also hold for any differences in level of education for both the two brand warmth conditions ²(4) = 1.597,

p = 0.809, as well as for the two levels of prestige ²(4) = 2.518, p = 0.641. To conclude,

these results show that the participants were equally distributed among the four conditions based on gender, age, and education.

5.4 Reliabilities and correlations

Table 1 presents the reliability, correlation, mean and standard deviation of the variables. As can be seen in Table 1, the reliability of the variables consumer-brand identification (α = .895), personal-identity expressiveness motive (α = .897) and service category involvement (α = .737) were all above the threshold of α > .70 and can be considered as good measurements. Although the variable engagement had a Cronbach alpha of .697, slightly below the threshold of .70, it can still be considered as a reasonable scale and will be used for further analyses.

Table 1 also presents the relevant correlations between the variables. As expected, brand warmth is positively related to (both types of) engagement r (267) = .282, p = .000. Unexpectedly, results show that brand prestige is not significantly related to the user’s

(27)

26

engagement in terms of comments r (267) = .114, p =.049. Furthermore, and as expected, both brand warmth and prestige have a significant and positive relationship with consumer-brand identification, respectively r (267) = .255, p = .000; r (267) = .137, p = .021. In addition, results showed that the motive to express personal identity on social media was positively related to both engagement and consumer-brand identification, respectively r (267) = .382, p = .000; r (267) = .236, p = .000. Additionally and as expected, consumer-brand identification was positively related to engagement r (267) = .482, p = .000.

Additionally, what became apparent is that the service category involvement has a positive significant relationship with the personal-identity expressiveness motive r (267) = .194, p = .002, and the highest level of education of participants was negatively related to the overall level of engagement r (267) = -.165, p = .009, as well as likes and comments

separately. Moreover, the Instagram usage also shows positive correlation with the variables brand warmth, consumer-brand identification, personal-identity expressiveness motive and service category involvement. Respectively, r (267) = .139, p = .024; r (267) = .135, p = .029;

r (267) = .192, p = .003 and r (267) = -.173, p = .005.

The r indicates the correlation coefficient that measures the strength and direction of a linear relationship between two variables, and values from minus one to plus one. The value zero until the absolute value .3 indicates a weak linear relationship, and values above the absolute value .7 indicate a strong linear relationship.

Table 2: Descriptives and correlations between the variables (Cronbach’s Alphas on diagonal)

M SD 1 2 3 4 5 6 7 8 9 10 11 1 Genderª 1.670 0.472 - 2 Educationb _4.050 _1.116 _.123 _- 3 Insta usage 3.120 1.453 .277** .063 - 4 Prestige 0.490 0.501 -.045 .004 -.025 - 5 Brand warmth 0.480 0.501 .120 .059 .139* .021 - 6 Engagement (likes) 3.280 1.841 .134* -.165** .105 .105 .332** - 7 Engagement (comments) 1.770 1.157 .007 -.174* -.059 .114* .124* .481** - 8 CBI 3.999 1.236 -.007 -.120 .135* .137* .255** .490** .303** (.895) 9 Engagement (total) 2.569 1.372 .134* -.165** .048 .112 .282** .915** .787** .482** (.697) 10 PI- expressiveness motive 3.001 1.518 .087 -.020 .192** .054 .070 .334** .307** .236** .382** (.897) 11 Service category involvement 5.025 1.053 .080 .110 .173** .093 .058 .107 .063 .081 .088 .194* (.737) Note : N = 267 * p<.05, **p<.01 ª 1 = Male, 2 = Female

b_{1 = Primary School, 2 = High School, 3 = Vocational Education/Trade School,}

(28)

27

5.5 Hypotheses testing

5.5.1. Mediation of Consumer-brand identification. After removing outliers and missing value cases, data from 262 participants was used for further analyses to test the hypothesis. In order to analyze whether consumer-brand identification mediates the relationship between brand warmth and consumer engagement, four ordinary least square regression analyses have been conducted. The outcomes of the four regression models are presented in Table 3. The constants in this table shows the values of consumer-brand

identification and engagement, when the variable brand warmth and the control variables are not present. In these regressions, there is controlled for the level of Instagram usage of consumers, the involvement of the consumers in the category of flying and brand prestige. Figure 2 shows a visual representation of the outcomes.

In order to argue for mediation, there must be a positive relationship between brand warmth and consumer engagement. As predicted, Table 3 Model 1 showed that brand warmth has a positive effect on consumer engagement (β = .267, p < .01). This indicates that when a picture posted by a brand shows that the brand is very loveable; the consumer will show more engagement in terms of likes/comments. The R² = .070 in the table indicates that in this model, 7.0 percent of the variance in engagement is explained by the brand warmth that is shown in the picture.

Furthermore, Model 2 of Table 3 presented with an explained variance of 5.6 percent, that brand warmth positively influences the consumer-brand identification of the consumer (β = .228, p < .01, R² = .056). This indicates that the higher the emotional appeal in the picture, the more a consumer will identify with this brand.

As mentioned in the literature review, consumer-brand identification often leads to consumer engagement. As becomes apparent in this study, results show indeed with an

explained variance of 22.3 percent, that consumer-brand identification has a positive influence on consumer engagement (β = .519, p < .01, R² = .223). This means that when the consumer-brand identification increases with a value of one, the level of overall engagement is expected to increase with .519.

Finally, Model 4 with 24.8 percent explained variance shows that the effect of brand warmth still significantly affects overall engagement (β = .165, p < .01), while the effect of consumer-brand identification also remained significant (β = .435, p < .01, R² = .236). This verification of the predictions provides evidence for the partial mediation model, supporting

(29)

28

relationship between brand warmth and the average level of engagement, but it does also mediate the relationship between brand warmth and engagement in terms of likes and

comments separately, respectively (β = .549, p < .05; β = .527, p < .01, R² = .360, β = .127, p < .05; β = .222, p < .01, R² = .200). These results can be found in table 5.

Table 3: Regression results on the mediation effect of CBI for all 4 models

Model 1 Model 2 Model 3 Model 4

Dependent variable

Engagement CBI Engagement Engagement

Coefficient SE Beta Coefficient SE Beta Coefficient SE Beta Coefficient SE Beta

Constant 1.687 .414 3.127 .379 .193 .427 .425 .250 Instagram usage .000 .058 .000 .084 .053 -.029 .053 -.031 -.045 .053 -0.47 Service category involvement .082 .078 .063 .046 .072 .039 .066 .072 .051 .061 .071 .047 Prestige .282 .164 .103 .315 .150 .126* .116 .152 .042 .123 .149 .045 Brand warmth .732 .165 .267** .571 .151 .228** .452 .153 .165** CBI .519 .061 .474** .477 .062 .435** R² .070 .056 .223 .248 Note : N = 267 * p<.05, **p<.01

5.5.2 Moderated mediation of brand prestige. After supporting the hypothesis that consumer-brand identification partially mediates the relationship between brand warmth and consumer engagement, there could be tested whether this mediation was moderated by brand prestige and the personal-identity expressiveness motive. Hypotheses 2, 3 and 4 were tested using ordinary least squares (OLS) regression with the conditional process modeling

(PROCESS) program for SPSS (Hayes, 2017). Model 22 has been used to test the interaction between X (brand warmth) and W (prestige) on M (consumer-brand identification) and the three-way interaction between X, W and M in affecting Y (engagement). In addition, this model is used to test the interaction effect between M (consumer-brand identification) and Z (personal-identity expressiveness motive) on Y (engagement). As mentioned, Model 22 has been conducted three times to investigate any differences in results between the two types of engagement separately, and the overall average level of engagement.

Table 4 presents the regression results for the moderated mediation of brand prestige and the personal-identity expressiveness motive with the average level of engagement as dependent variable. In these regressions, results have been controlled for the effects of Instagram usage and service category involvement. a1 in this table indicates that consumer-brand identification is estimated to increase 0.634 units as consumer-brand warmth changes for 1 unit. This effect is significant (p < .01).

(30)

29 Hypothesis 2 stated that the prestige of a brand moderates the relationship between brand

warmth and engagement, such that high levels of brand prestige will increase the effect of brand warmth on the user engagement. With an explained variance of 37.9 percent, the results indicate that the effect of showing brand warmth in a picture on consumer engagement is conditioned by the prestige of a brand, as evidenced by a statistically significant interaction (c’3 = .575, p <.05, R² = .379). A closer examination of this conditional effect indicates that only the direct effect is conditioned by prestige. The direct effect means the effect of brand warmth on customer engagement, without the mediation of consumer-brand identification. As the value zero is within the confidence interval of low prestige brands, and outside the

confidence interval of high prestige brands, the direct relationship between brand warmth and consumer engagement appears as significant only for high prestige brand (effect = .201, SE = .199) compared to the low prestige brand (effect = .776, SE = .198). This confirmation of predictions provides evidence for the moderation effect of brand prestige on the relationship between brand warmth and the overall level of user engagement, supporting hypothesis 2. A visual representation of this effect can be found in Figure 2b.

Hypothesis 3 stated that brand prestige moderates the indirect, and thus the mediated

relationship between brand warmth and consumer engagement by consumer-brand

identification, such that high levels of brand prestige will result in a higher effect of brand warmth on consumer-brand identification. However, results indicate that the level of brand prestige appears to have no effect on the mediated relationship between brand warmth, consumer-brand identification and user engagement (a3 = -.141, ns) and therefore reject

hypothesis 3 with an explained variance of 8.80 percent (R² = .0880). Hence, the effect of

brand warmth in pictures on the consumer-brand identification is the same for both high and low prestige brands.

5.5.3 Moderated mediation of personal-identity expressiveness motive. Similarly with the model used to test hypothesis 2 and 3, hypothesis 4 that states that personal-identity expressiveness motive moderates the mediated relationship of consumer-brand identification on brand warmth and user engagement, was also tested using model 22 with ordinary least squares (OLS) regression with the conditional process modeling (PROCESS) program for SPSS (Hayes, 2017).

The results in Table 4 indicate that the effect of consumer-brand identification on engagement is conditioned by the level of the personal-identity expressiveness motive, as evidenced by a statistically significant interaction (b3 = .091, p <.01), supporting hypothesis

Driving engagement by generating consumer-brand identification : how can service brands benefit from the shift to visual content in social media marketing?

Driving engagement by generating consumer-brand identification

Table of contents