Are influencers more credible advertisers of luxury brands?

(1)

Are influencers more credible advertisers of

luxury brands?

The extent to which social media marketing impacts consumer’s purchase

intentions and the effects of brand type and source credibility

University of Groningen Faculty of Economics and Business

Master`s Thesis MSc. Master Management

January 13th, 2020 Supervisor: dr. J.C. Hoekstra Second supervisor: prof. dr. J. van Doorn

(2)

1

Abstract

This study investigated the impact of social media marketing on purchase intentions. Besides, the effect of the moderator brand type and the mediating role of source credibility were analyzed. Moreover, this study controlled for the attitude toward the brand and attitude toward the post. An online experiment (N = 170) was conducted, and a 2 (company profile vs. influencer profile) x 2 (non-luxury brand vs. luxury brand) between-subjects design was used. First, it was found that there is no difference between influencer profiles and company profiles condition on the effect of social media marketing on purchase intentions. Second, similar results were found for the effect of social media marketing on source credibility. Third, contributing to the existing literature, source credibility was found to have a positive effect on purchase intentions. Fourth, it was found that brand type does not significantly moderate the effect of social media marketing on either purchase intentions or source credibility. Moreover, there was no significant difference between the luxury brands or the non-luxury brands condition on the effect of social media marketing on source credibility and respectively, purchase intentions. Besides, source credibility does not mediate the effect of social media marketing on purchase intentions. Lastly, results and implications are discussed.

Keywords: social media marketing, purchase intentions, brand type, luxury brands, non-luxury

(3)

2

Preface

This represents the last step of my master`s degree in Marketing Management at the University of Groningen. I was lucky enough to have a thesis topic to which I have enjoyed working on for the last 5 months and namely, the topic of influencer marketing and luxury brands. It has been a year full of ups and downs with plenty of hectic days (and mostly nights) and regular 10-hour long study sessions at the UB or Zernike library. Only now I realize how much I will miss these days and especially, the beautiful city of Groningen.

(4)

3

1. Introduction

Advertising plays a vital role in our everyday lives, and it is expected to advance as more and more consumers are shifting their attention from traditional advertising media to online media (Kitchen & Uzunoglu, 2015). Internet advertising will reach a growth of 4.7% in 2019 regarding the total amount of money invested in advertising worldwide, and forecasts show a 4.6% growth in 2020 and 2021 (Zenith Media, 2019). The same study forecasts that online advertising will reach the 50% mark of the total global advertising expenditures for the first time in 2021. This growth can be explained by the increasing social media usage of our days, especially among young adults aged 18 to 29 years old, of which 90% use social media, compared with a 12% usage in 2005 (Perrin, 2015).

Social media has made creating, publishing, and sharing content with a large community that has similar interests more accessible than ever. Companies have recognized the persuasion power of online opinion leaders (Uzunoglu and Kip, 2014), especially on social media platforms such as Instagram. Thus, this led to the creation of a new sub-category of social media marketing known as influencer marketing. Literature has defined this relatively new concept as the art and science that engages influential people with their community of followers and to which the brand message can be sent in the form of sponsored content (Sammis, Lincoln & Pomponi, 2016). These influential people are known as creators of content that provide their followers with insights about their lifestyle, experiences, and opinions and which are often compensated by brands to endorse their products among their vast communities of followers.

Nowadays, brands are shifting towards placing their products in the online environment by advertising on their social media marketing channels or with the help of online influencers (Tutaj & Van Reijmersdal, 2012). According to the same scholars, this engages the persuasion knowledge at a later stage and thus, leads to higher effectiveness of the advertisement. Business firms that choose to advertise on their social media marketing channels (e.g. company profiles) are integrating marketing activities at a more favorable cost-benefit ratio than before (Kim & Ko, 2012). Firm-created content is seen as a fundamental component of the promotion mix and managers use it to increase consumers' brand perceptions and bring new customers on board (Schivinski & Dabrowski, 2016).

(7)

6 while influencers share their product-related posts with their community, providing a an advantageous relationship for both the brand and the influencer (De Veirman, Cauberghe and Hudders, 2017). The confirmation that influencer marketing is becoming an established marketing technique for the whole industry comes from luxury brands. These brands, known for their exclusivity and association with the high society, were the last adopters of social media marketing as there were concerns about losing their brand integrity (Pentina et al., 2018).

However, several studies show otherwise. Consumers are significantly impacted by forms of electronic word-of-mouth, such as influencer marketing, as they perceive it as a more credible source that shares an honest opinion about the product, rather than posts by companies, where the intent to sell is revealed (Shuqair, Cragg and Mitchell., 2016). Moreover, Cheah et al. (2015) found that consumers are more likely to buy luxury brands when significant social influence is exercised, as it is the case with influencer marketing. Over the years, scholars have tried to understand the effects of social media on brands and consumers` purchase intention by researching related topics such as online influencers (Pornpitakpan, 2004), luxury brands perceptions (Lee & Watkins, 2016) and source credibility (Lim, Radzol, Cheah, Wong, 2017).

A study by Lee & Watkins (2016) has found that respondents exposed to influencer content had higher purchase intentions and more positive luxury brand perceptions, compared to the control group. Moreover, social interaction was found to moderate the relationship. By para-social interaction, we refer to a one-way relationship between an individual and a ‘media persona,’ in which the individual invests time, interest, and energy while the other party is not even aware of its existence (Horton and Wohl, 1956). This proves that consumers will follow their favorite internet ‘friends’ and when these bloggers form partnerships with brands, their followers are more likely to become followers of the brand.

(8)

7 difficulty of assessing the credibility of the source in the online environment (De Veirman et al., 2017). Consequently, this leads to our research question: are influencers more credible advertisers of luxury brands (vs. non-luxury brands), compared to brands themselves?

This study aims to shed light on this question by bringing forward evidence and managerial implications for professionals involved in the industry of luxury brands. Its goal is to help brands and managers two-fold. First, we aim to show whether the adoption of social media marketing has a significant impact on purchase intentions, and secondly, whether influencers are credible sources that lead to higher purchase intentions when a luxury product is endorsed, compared to a non-luxury product. This paper is structured as follows. First, a literature review and the developed hypothesis are described. Second, the methodology is presented, which includes the procedure of the experiment, followed by the plan for data analysis. Hereafter, the results of the analyses are described, followed by a discussion, conclusion and implications.

2. Literature review

2.1 Conceptual model

This study investigates the effect of social media marketing (company profiles vs. influencer profiles) on purchase intentions. First, it is expected that the effect of social media Figure 2.1: Conceptual Model

(9)

8 marketing on purchase intentions will be larger in the influencer condition, compared to the company condition. Second, it is examined whether social media marketing affects the measured variable source credibility and it is expected that the effect of social media marketing will be larger in the influencer condition compared to the company condition. Third, source credibility is expected to affect purchase intentions positively. Fourth, the effect of the moderator brand type (luxury vs. non-luxury brands) on the relationship between social media marketing and source credibility will be analyzed. It is expected that the effect of influencer profiles will be larger in the luxury condition, than the effect of company profiles, compared to the non-luxury condition. Afterward, the effect of the moderator brand type on the relationship between social media marketing and purchase intentions will be analyzed. Similar expectations are predicted. Lastly, it is expected that source credibility mediates the effect of social media marketing on purchase intentions. Additionally, the effect of control variables will be taken into account by measuring the attitude toward the brand and the attitude toward the post of the respondents. Consequently, the conceptual model presented in Figure 2.1 was developed.

2.2 Company profiles versus influencer profiles

According to a study by Simply Measured (2004), among the regular users of Instagram, 86% of the best-ranked brands in the world have official accounts on this social media platform that regularly advertise their brands and products. Their end-goal is to engage with current, future, or potential customers and raise brand awareness. Therefore, brands see this platform as an opportunity to come into contact with their clients (Michaelidou et al., 2011). Content created by firms on their social media marketing channels is known as a type of advertising which follows the company`s marketing strategy (Schivinski & Dabrowski, 2014). The advantages that company profiles or official brand accounts have are numerous: marketers can receive feedback regarding their customers’ expectations or needs and resolve them formally or informally (Erkan, 2015). According to the same scholar, brands can engage in discussions to prevent current issues from becoming even more prominent and they can also improve communication with the client by providing accurate and up-to-date content.

(10)

9 lead to higher purchase intentions. Moreover, the same study has found that 85% of these customers perceive this type of content as ‘far more trustworthy than posts by the product owner or advertiser’ which indicates that customers perceive social media influencers as a more credible and honest source of electronic word-of-mouth compared to posts made by official company accounts (Shuqair et al., 2016, p. 16)

As it is a relatively new form of marketing, there are several definitions of influencer marketing, such as the one provided by the leading influencer marketing company, TapInfluence (2017). They define influencer marketing as a new marketing type that is focused on opinion leaders and in which businesses ‘inspire, hire, or pay influencers to get the word out’. Other scholars, such as Sammis, Lincoln & Pomponi (2016), define influencer marketing as the art that engages persuasive people into sharing branded content with their followers on a social media platform. According to the same researchers, influencer marketing is similar to celebrity advertisements because people tend to have more trust in an individual that they admire or that they can relate with. This success is due to the fact that consumers can create and distribute brand-related content that shows their product preferences to their peers and acquaintances through social media such as Instagram and Facebook, which makes the most reliable tool for electronic word of mouth (De Veirman et al., 2017).

Influencers are highly related to electronic word-of-mouth (eWOM) as consumers interpret influencers` posts on social media platforms, as a form of eWOM which is more credible than traditional advertising (Abidin, 2016). Moreover, influencers are persuasive individuals that acquired an active community of followers by creating and disseminating content through blogging or vlogging on social media platforms (De Veirman et al., 2017). Compared to other types of social media marketing, influencers are perceived as easier to relate to because they tend to share the content of their private lives, which makes them be perceived as more trustworthy and credible (De Veirman et al., 2017).

2.3 Purchase intentions

(11)

10 plan that an individual has towards acquiring a particular brand or product. According to Lim et al. (2017), influencer marketing can boost consumer`s purchase intentions as influencers are perceived as more genuine in the presentation and dissemination of the message towards their followers. Also, according to the same study, consumers are more likely to be persuaded by an influencer`s recommendation due to the rapport that was built with the users and hence, have increased purchase intentions. Besides, scholars have found that purchase intention is positively impacted by e-WOM (e.g. influencer marketing), as customers can read reviews about the product that they are interested in, right at the point-of-purchase. Moreover, Pornpitakpan (2004) found a direct effect between the effect of social media influencers and consumer`s purchase intentions.

In contrast, social media marketing channels owned by companies (e.g. company profiles) were found as too intrusive by internet users, which leads to avoidance responses (Benes, 2018). Similarly, other scholars have found that intrusiveness negatively affects purchase intentions (van Doorn & Hoekstra, 2013) and thus, influencer profiles are expected to have a higher impact on purchase intentions, rather than when a company advertises the same product. Consequently, we propose the following hypothesis:

H1: For influencer profiles, the effect of social media marketing on the purchase intention

is larger than for company profiles.

2.4 Source credibility

Source credibility has been analyzed by prior research (Ohanian, 1990) and identified as being composed of three underlying dimensions: attractiveness, expertness, and trustworthiness. While expertness refers to a person`s level of sufficient knowledge about a particular topic, trustworthiness refers to the perception that knowledge is unbiased. Lastly, attractiveness refers to how the endorser`s appearance affects the viewers. As other researchers (Till and Busler, 1998) have suggested that the effect of attractiveness is not powerful enough to make an impact on purchase intentions, this dimension was not taken into account in this study.

(12)

11 followers (Lim et al., 2017; Metzger et al., 2013).

When a user follows an influencer, he expects the influencer`s opinion to be the result of his own experience, unbiased even if a company pays the post or not (Hall, 2016). On the other hand, if the promotion doesn`t seem sincere enough as a result of an unsuccessful alignment between the brand and the influencer`s industry, the consumers may not respect the influencer`s opinion which leads to adverse outcomes on the brand and the influencer as well (Hall, 2016). Thus, if an influencer from the fashion industry creates fashion-related content, users will perceive him as a source of expertise and trustworthiness. Therefore, an influencer endorsing fashion would be perceived as an expert rather than a fashion influencer endorsing a product that is not aligned with his niche. If an influencer is perceived as an expert, the consumer will view the shared content as accurate information about the brand, which enhances the trust levels of the shared message (Fogg, 2002). Social media influencers became a reliable source of information for internet users, which led to numerous PR campaigns coming from organizations whose goal was to strategically influence their audience (Mackiewicz, 2010; Miller, 1989). Even with these promotional campaigns, strong positive effects have been found between the credibility of an influential advertiser and consumer`s attitudes toward the brands, resulting in positive attitudes toward the products (Schenk and Döbler, 2002).

When an influencer is initiating the communication of the message, it is perceived as more authentic and credible, compared to when an advertiser (e.g. company profile) communicates it, as the act of following an influencer is a conscious choice (De Veirman et al., 2017). Besides, internet users are more likely to perceive an influencer`s recommendation as more trustworthy than a brand`s recommendation as consumers identify at a personal level with influencers (Schouten, Janssen & Verspaget, 2019). In the same way, when a product is endorsed in an environment that resembles real-life settings (e.g influencers that have already tried the product), the perceived credibility will be higher, compared to other forms of advertising where the monetary incentive is obvious (e.g. company profiles) (Schouten et al., 2019; Shuqair et al., 2016). As such, source credibility is expected to be higher in the influencer condition, compared to the company profile condition and the following hypothesis is developed:

H2: For influencer profiles, the effect of social media marketing on source credibility is

(13)

12

2.5 The effect of source credibility on purchase intentions

Literature suggests that a more credible source is more efficient in getting attention and increasing recall levels (Sternthal, Philips and Dholokia, 1978). Thus, a more credible source is more likely to produce positive changes and attitudes towards a particular product, which results in extensive financial efforts coming from companies (Lafferty and Goldsmith, 1999). Moreover, source credibility was found to have a positive relationship with e-WOM, such as social media marketing, when consumers put more emphasis on the credibility of the source rather than the type of social media that is being used (Hui, 2017). Consequently, as e-WOM represents a more credible source, due to its unbiased perspective, it was found to radically affect consumer`s purchase decisions (Cheung, Lee, and Rabjohn, 2008). Moreover, trustworthiness and expertise, two of the dimensions of source credibility as identified by Ohanian (1990), are positively related to purchase intentions (Pornpitakpan, 2004).

Other studies support this view and approach source credibility from the brands` point of view and find that as a brand is perceived as more credible, due to being more favorable, unique or strong, the more it impacts the consumer`s purchase intentions (O’Cass & Lim, 2001). This is also the case with Wang & Yang (2010), who found that there is a positive relationship between the credibility of the brand and purchase intentions. They proved that as a brand becomes more credible, the higher the chances that consumers will consider it, which leads to higher purchase intentions. Therefore, the following hypothesis is developed:

H3: Source credibility positively affects purchase intentions. 2.6 Type of brands and social media

In this study, two types of brands will be assessed according to the expectations, and such, the terms luxury brands and non-luxury brands will be used. According to Vigneron and Johnson (2004), which have focused on how the perceived luxuriousness of a particular brand makes the consumers express themselves towards an ideal self or trying to get as close as possible to that self by using a specific brand. In the same paper, luxury is seen as a significant factor of differentiation for a product or a brand, and even as a factor that defines consumption across cultures, enriching consumers' choice.

(14)

13 brings self-esteem to the owner. In line with their study, the brand Gucci was chosen for this research. On the other hand, in defining the non-luxury goods, the concept provided by Truong, Simmons, McColl & Kitchen (2008) will be used. In their view, store brands are not associated with luxury goods and therefore, the brand Bershka was chosen for this research.

This study focuses on official social media profiles owned by luxury brands, non-luxury brands, and influencer profiles. However, the literature regarding this topic is still scarce. Still, prior research about brand profiles has analyzed the effect of official social media accounts of luxury brands on social media platforms. This can be seen in the framework developed by Kim & Ko (2012). Their research includes five interaction points between social media marketing and luxury brands: entertainment, interaction, trendiness, customization, and word-of-mouth. Entertainment refers to the pleasure of exploring the brand`s page, while interaction refers to holding conversations and sharing knowledge on the product`s page. On the other hand, trendiness is the inclusion of the newest information about the brand and customization refers to the customized information search. Lastly, word-of-mouth is the ability to share the information found on the luxury brand`s page with other consumers, outside of the social media platform.

Scholars suggest that luxury brands were reluctant to the adoption of new tools such as social media marketing, but they have more recently started to be aware of the benefits (Cailleux, Mignot & Kapferer, 2009). Luxury brands use social media to gain attention and to stimulate the eagerness towards their products, but also to obtain feedback directly from the customers (Kim & Ko, 2012).

2.7 The moderating role of brand type

(15)

14 consumer`s attitudes toward luxury brands (Tak & Pareek, 2016) which ultimately leads to higher purchase intentions (Bian & Forsythe, 2012).

Other scholars have suggested that the type of brand has a positive effect on consumer`s purchase intentions (Kim & Ko, 2010). In the same way, luxury brands enhance purchase intentions (Bian & Forsythe, 2012) due to the parasocial interaction with the influencer (Lee & Watkins, 2016). Para-social interaction refers to an imaginary relationship of friendship and the illusion of an interpersonal connection between a media persona and a regular user, which is one-sided (Tsiotsou, 2015; Horton & Wohl, 1956). The evolution of connectivity and media endorsers through social media platforms has brought the possibility of two-way contact between influencers and users. However, research has shown that the interaction is still one-sided (Stever & Lawson, 2013). The effect of para-social interaction has been linked with Instagram celebrities (Ward, 2016). Such an example would be a celebrity of our days (e.g. Selena Gomez), that regularly posts personal content of her everyday activities on Instagram. In this way, her community of followers feels entertained, updated and can share content with other peers (word-of-mouth), even if the interaction is one-sided, following the framework proposed by Kim & Ko, 2012). Therefore, influencer marketing is in line with the findings of the para-social interaction (Tsiotsou, 2015; Horton & Wohl, 1956) and the interaction between social media and luxury brands (Kim & Ko, 2010).

Lee & Watkins (2016) found that after consumers were exposed to content provided by influencers, that included luxury products, their purchase intentions were higher. Moreover, the literature suggests that one compares himself upward with peers that he considers of higher value than his own person (e.g. social media influencers) and thus, para-social interaction levels increase and so does purchase intentions for luxury brands (Lee & Watkins, 2016). On the other hand, Calvo-Porral & Lévy-Mangin (2017) have used store brands (e.g. non-luxury brands) as a moderator in their research. Their findings show that this type of brands leads to lower purchase intentions, for high-quality seeking customers, as compared to manufacturer brands (e.g. luxury brands). Therefore, the relationship between social media marketing and purchase intentions would be higher when a luxury brand is endorsed, compared to a non-luxury brand, in the influencer condition. Consequently, the following hypothesis is developed:

H4: For luxury brands, the effect of influencer profiles on purchase intentions, compared

(16)

15 Section 2.4 suggests that an influencer profile has a higher impact on source credibility, rather than a company profile. By taking into account the outcomes of para-social interaction described in the previous section, the effect of social media marketing on source credibility is expected to also vary according to the type of brand.

These results are expected as, influencer marketing, a form of native advertising (Evans, 2017), creates an impression over the consumer`s perception that can be linked with the Persuasion Knowledge Theory. The theory developed by Friestad and Wright (1994) states that people learn when different persuasion agents such as advertisers or salespeople are trying to persuade them, and their behavior changes accordingly to their defense system. This behavior requires time and acquiring of cumulative knowledge about persuasion attempts. When the consumer can recognize these attempts, it becomes possible for the user to guard himself against unwanted persuasion. In the case of social media influencers, the persuasive motive is less obvious and thus, the defense system occurs at a later time (Tutaj & van Reijmersdal, 2012). Additionally, according to Evans et al. (2017), consumers may realize that the post that they are watching is, in fact, a persuasive attempt only when a disclosure is present and when there is ‘retrospective memory of disclosure’, compared to company posts where the intent to advertise is obvious. Furthermore, this view is supported by the findings of Isaac & Grayson (2017), which state that if the persuasive agents (e.g. influencers) use a credible tactic, the persuasion knowledge model leads to increased levels of credibility of the source.

(17)

16

H5: For luxury brands, the effect of influencer profiles on source credibility, compared to

company profiles, is larger than for non-luxury brands.

2.8 The mediating role of source credibility

The framework by Ohanian (1990), which contains three dimensions of source credibility (expertise, trustworthiness, and attractiveness) can be associated with the characteristics of social media influencers (Djafarova & Trofimenko, 2019). By following the same framework of Ohanian (1990), Pornpitakpan (2004) has shown that online endorsers affect consumer`s credibility of the source which inevitably leads to higher purchase intentions. Additionally, similar results were found by Wang and Yang (2010), which state that brand credibility has a positive effect on purchase intentions. Moreover, Anridho & Liao (2013) have used endorser credibility as a mediator in a similar context. In the same way, Chan et al. (2013) have found a positive correlation between online endorsers, the credibility of the endorser and purchase intentions. Nevertheless, other scholars have even argued that source credibility acts as a mediator in the context of online advertising, as the effectiveness of the endorsement depends on whether the source is perceived as credible or not (La Ferle & Choi, 2005), which ultimately affects purchase intentions (Pornpitakpan, 2004). Therefore, it is expected that social media marketing increases consumers` source credibility, which, in turn, will have an impact on their purchase intentions. By connecting these arguments, the following hypothesis was developed:

H6: Source credibility mediates the relationship between social media marketing and

purchase intentions.

3. Methodology

3.1 Research design

(18)

17

Table 3.1: Overview of the four conditions and the number of respondents

Thus, we manipulated the independent variable social media marketing and the moderator brand type, while the mediator source credibility and the dependent variable purchase intentions were measured. Additionally, the following control variables were measured: attitude towards the brand, attitude towards the ad and demographics. Therefore, four experimental conditions were created as shown in Table 3.1.

3.2 Procedure and manipulation check

A pre-test was conducted to make sure that the experiment is viable and that the manipulation worked as intended. The 43 participants were exposed to the conditions, and afterward, their perception about the brand type and the social media marketing manipulation has been checked. To make sure that the respondents have the same understanding of the kind of brand to which they were exposed, the definition of luxury goods by Vigneron and Johnson (2004) was provided. As such, they were asked about their perception of the brand (luxury vs. non-luxury), by using the following self-made question: “Having in mind the text that you have just read, to what extent do you (dis)agree with the following statement?”. The provided statement was: “The brand I just saw is a non-luxury/luxury brand”. Respondents were able to answer by using a seven-point differential semantic scale.

To check the social media marketing manipulation, respondents were also asked about their perception by using a single self-made question, as follows: “Do you perceive the Instagram post that you have been exposed to as a profile owned by a company/influencer?”. There were two possible answers – company profile and influencer profile. Additionally, small grammar

(19)

18 mistakes were corrected and the possibility to go back to the previous questions was added, according to the feedback received from the participants.

An Independent T-Test was used to measure whether the manipulations differed significantly and whether they were perceived in an intended way or not. The variable brand type has been tested and results have shown that there is a significant difference (p < 0.001) in brand perception between the luxury and non-luxury condition (Mluxury = 5.90, SD = 1.70; Mnon-luxury =

3.59, SD = 2.01). Similar results were also obtained for the manipulation check of the final experiment, (Mluxury = 6.30, SD = 1.11; Mnon-luxury = 4.21, SD = 2.03). Therefore, this means that

the manipulation has worked as intended and the respondents have perceived Gucci as a luxury brand and Bershka as a non-luxury brand in both the pre-test and the final experiment.

The same test was used to measure whether the manipulation worked as intended with the other variable, social media marketing. It has been tested and results have shown that there is a significant difference (p < 0.001) in profile perception between the influencer profile and the company profile condition (Minfluencer = 1.35 SD = 0.48; Mcompany = 1.70, SD = 0.47). Similar results

were also obtained for the manipulation check of the final experiment, (Minfluencer = 1.18 SD = 0.38;

Mcompany = 1.62, SD = 0.49). Therefore, this means that the manipulation has worked as intended

and the respondents have perceived the profile ‘cristiano’ as a profile owned by an influencer and the profile ‘gucci’/’bershka’ as profiles owned by brands in both the pre-test and the final experiment.

The actual experiment was realized on the Qualtrics platform, and it has been shared among the researcher`s contacts on social media platforms such as Facebook and Instagram, due to constraints of time. A lottery was introduced to make the participation more appealing and to encourage respondents to answer to the best of their capabilities. Thus, participants had the chance to win a personalized Dopper water bottle. The winner was chosen by using the website random.org.

In the introduction of the survey, participants were informed about the general purpose of this research and their anonymity rights. Additionally, they were informed that they need an Instagram account and have to be over 18 years old to participate. The researcher`s email was provided to the respondents for further questions. At the bottom of the page, they were asked to confirm that they have read and comprehended the briefing.

(20)

19 clicked on yes, then they would have been redirected to the end of the survey.

Secondly, each participant was randomly allocated to one of the four conditions created for the experiment, and they were asked to imagine that they are browsing on Instagram and that they are following the profile that they are being exposed to. Additionally, to make sure that the participants would notice the manipulation, an explanatory text about the brand/influencer and the endorsed product was provided.

Thirdly, respondents were asked to answer questions regarding the attitude toward the brand, attitude toward the post, source credibility, and purchase intentions. All of the questions were identical in every condition. The following question was an attention check question in which the respondents were asked about the type of product that they have been exposed to. If they chose the right answer (a pair of shoes), then it was concluded that they had paid attention to the survey. Oppenheimer, Meyvis, and Davidenko (2009) proved that the inclusion of an attention check question improves statistical power and the reliability of the analysis.

Lastly, participants were asked to fill in their age, gender, and education level. In the debriefing section, the subjects were informed that the Instagram post was fictive and none of the persons/brands involved in the experiment have shared these posts on Instagram. When the respondents finished answering the questions, they were thanked for participating and were invited to fill in their email addresses to receive the results of the study or to find out if they were the winners of the lottery.

3.3 Stimuli

The respondents were first exposed to an explanatory text to make sure that the manipulation was noticed, according to the condition that was presented, as follows: “I kindly ask you to imagine that you are browsing on Instagram and the following post of an account that you follow shows up”. Please take your time to read the short description and look at the post as the next questions will refer to it”. Short description: Bershka/Gucci is a clothing brand known worldwide. On Instagram, their official account (profile) is called 'bershka'/’gucci’. In this post, they show a Bershka/Gucci shoe. In the condition where an influencer`s profile was present, the company name was replaced by Cristiano Ronaldo (‘cristiano’ on Instagram).

(21)

video-20 editing software Adobe Photoshop. According to Influencer Marketing Hub (2019), Cristiano Ronaldo is ranked as the number 1 influencer in 2019 regarding both the highest number of followers (over 160 million) and by being the most engaging influencer on Instagram. Therefore, we have chosen him for our study as he is known worldwide and has fans of both genders that are susceptible to his endorsements.

The caption was inspired by other influencers/brand posts and in the end, we decided to add a hashtag containing the name of the brand and a positive aspect of the product. The caption that resulted is:” New shoes from #gucci / #bershka! A must-have!”.

Therefore, for condition one, we have used the Gucci Logo and the influencer`s name (cristiano), while for the second condition, the influencer`s name was replaced by Gucci`s profile. For the third condition, we have used the Bershka Logo and the name of the same influencer, while for the fourth condition, we removed the name of the influencer and replaced it with Bershka`s profile (see Figure 3.1).

(22)

(23)

22

3.4 Measures

Participants were asked to answer questions regarding the independent variable (social media marketing: company profile vs. influencer profile), the dependent variable (purchase intentions), the mediator (source credibility), the moderator (brand type: luxury vs. non-luxury brands) and the control variables (attitude toward the brand, attitude toward the post and demographics).

Table 3.2: Construct measurements, items, factor loadings, and Cronbach`s alpha Construct (source) Items Factor loadings Cronbach alpha Source credibility (Ohanian, 1990)

How do you perceive the influencer/brand? • Dependable/undependable • Honest/dishonest • Reliable/unreliable • Sincere/insincere • Trustworthy/untrustworthy • Expert/not an expert • Experienced/inexperienced • Knowledgeable/unknowledgeable • Qualified/unqualified • Skilled/unskilled .69 .80 .91 .85 .87 .86 .87 .89 .89 .88 .96 Attitude toward the brand

(Spears & Singh, 2004)

How do you perceive the brand?

• Unappealing/appealing • Bad/good • Unpleasant/pleasant • Unfavorable/favorable • Unlikable/likable .93 .93 .96 .94 .95 .97 Attitude toward the post (Henthorne et al., 1993)

(24)

23

Purchase intentions How likely would it be for you to buy the endorsed product?

(Chandran & Morwitz, 2005)

• Highly unlikely/highly likely .94

How probable is it that you will purchase the endorsed product?

• Highly improbable/highly probable .98 .97

How certain is it that you will purchase the endorsed product?

• Highly uncertain/highly certain .95

What is the chance that you will purchase the endorsed product?

• No chance at all/very good chance .97

Note: All items have been measured by using a 7-point Likert scale 3.4.1 Source credibility

Source credibility was measured by using the framework proposed by Ohanian (1990), which contains three sub-dimensional scales with five items each. The sub-dimensions refer to physical attractiveness, trustworthiness, and expertise. The sub-dimension ‘physical attractiveness’ was removed because, in our experiment, only products were involved in the manipulation and also, because of prior findings, which suggest that this dimension is not relevant regarding the evaluation of celebrity endorsers (Till & Busler, 1998).

3.4.2 Demographics

Respondents were asked to mention their age (open question), gender and education level. The possible answers for gender were male/female, while possible answers for education level were no education, primary education, high school diploma, Bachelor`s degree, Master`s degree, Ph.D./Doctorate.

3.5 Participants

(25)

24 left us with a total of 170 respondents. The remaining participants were randomly assigned to one of the conditions of Table 3.1.

3.6 Factor analysis and reliability analysis

Before the factor and reliability analysis was conducted, the dataset was cleaned and prepared for the analyses. Unnecessary data from Qualtrics was removed and no outliers or missing values were found. We identified a total of 36 respondents that needed to be filtered out because they have either failed the attention check question or they were not Instagram users. According to Malhotra et al. (2010), it is recommended to check if the factor analysis is appropriate before proceeding. For the factor analysis to be appropriate, the value of the KMO-statistics must be ≥ 0.5, and Barlett`s Test of Sphericity must be significant with a p-value < than 0.05. Additionally, communalities must have values which are > than 0.4, indicating that the amount of variance in each variable is good enough (Malhotra, 2010). Furthermore, the reliability of the scales was assessed by using Cronbach`s Alpha, which should have a value higher than 0.6 to be considered reliable (Malhotra, 2010). The constructs were analyzed separately as follows:

First, the construct that contains five items, attitude toward the brand, was analyzed and factor analysis was found to be appropriate as the value of the KMO-statistics (0.89) was higher than 0.5 and Bartlett`s Test of Sphericity had a highly significant result (p < .001). Besides, communalities were checked and found to be over the value of 0.4. Regarding the reliability of the scale, it was found that the scale by Spears and Singh (2004) is reliable as Cronbach`s Alpha was higher than 0.6 (0.968), (see Table 3.2).

Second, the construct that contains six items, attitude toward the post, was analyzed and factor analysis was found to be appropriate as the value of the KMO-statistics was higher than 0.5 (0.804) and Bartlett`s Test of Sphericity had a highly significant result (p < .001). Communalities were checked and found to be over the value of 0.4. It was decided to remove the item ‘Easy to understand’ as the value of ‘Cronbach`s Alpha if item deleted’ showed that it would increase from a value of 0.85 to 0.87. Regarding the reliability of the scale, it was found that the scale by Henthorne et al. (1993) is reliable as Cronbach`s Alpha was higher than 0.6 (0.87), (see Table 3.2).

(26)

25 0.5 and Bartlett`s Test of Sphericity had a highly significant result (p < .001). Also, communalities were checked and found to be over the value of 0.4. Regarding the reliability of the scale, it was found that the scale by Ohanian (1990) is reliable as Cronbach`s Alpha was higher than 0.6 (0.96), (see Table 3.2).

3.7 Plan for analysis and assumptions check

Before performing the analyses, several assumptions have been checked for each chosen statistical analysis (see Table 3.3). Moreover, the estimated regression models for Hypotheses 3, 4, 5 and 6 are presented in Table 3.4.

Table 3.3: Overview of used statistical analyses for each hypothesis

Hypotheses Statistical analysis H1: For influencer profiles, the effect of social media

marketing on purchase intention is larger than for company profiles.

One-Way ANCOVA

H2: For influencer profiles, the effect of social media

marketing on source credibility is larger than for company profiles.

One-Way ANCOVA

H3: Source credibility positively affects purchase

intentions.

Multiple regression

H4: For luxury brands, the effect of influencer

profiles on purchase intentions, compared to company profiles, is larger than for non-luxury brands.

Regression (Hayes` model 1)

H5: For luxury brands, the effect of influencer

profiles on source credibility, compared to company profiles, is larger than for non-luxury brands.

H6: Source credibility mediates the relationship

between social media marketing and purchase intentions.

(27)

26 assumption that there is linearity between the covariates attitude toward the brand and attitude toward the post and the dependent variable purchase intentions (H1) has been checked. After visually assessing the scatterplot, a linear relationship has been found between each covariate and the dependent variable for each intervention type. Similar results were also found with source credibility as the dependent variable (H2).

Second, the assumption of homogeneity of regression slopes has been checked. The tests of between-subjects effects showed that the relationship between the independent variable social media marketing and the covariate attitude toward the brand is not significant, F(1,00) = .00, p = .99. Additionally, the relationship between social media marketing and attitude toward the post is not significant, F(1,10) = .51, p = .48. As the relationship between social media marketing and both of the covariates is not significant, the assumption of homogeneity of regression slopes has not been violated. Similar results were found for H2, the tests of between-subjects effects showed that the interaction between the independent variable social media marketing and the covariate attitude toward the brand is not significant, F(1,164) = .00, p = .95. Moreover, the interaction between social media marketing and attitude toward the post is also not significant, F(1,164) = 1.31, p = .25.

Third, the assumption of normality has been checked. Therefore, the Kolmogorov-Smirnov (p = .20) and Shapiro-Wilk test (p < .001) showed that this assumption has been violated, and thus, the standardized residuals are not normally distributed (H1). The assumption of normality has also been violated in the case of H2, as both the Kolmogorov-Smirnov (p < .001) and Shapiro-Wilk test (p < .001) were significant. However, when the sample is large enough (> 50), significant results can be obtained even if there is a slight deviation from normality (Ghasemi and Zahediasl, 2012).

Fourth, the homoscedasticity has been checked by plotting the standardized residuals against the predicted values, and as no discernable pattern has been found, the assumption has not been violated. Besides, the assumption of homogeneity of variances has not been violated (H1) as Levene`s Test is not significant (p >.05). Similar results were obtained for H2 and thus, the assumption of variances has not been violated as Levene`s Test is not significant (p > .05). No outliers were found as no cases where the standardized residuals with values higher than ±3 standard deviations were discovered in both of the hypotheses (Field, 2013).

(28)

27 and attitude toward the post. Several assumptions were checked. First, the assumption regarding the independence of observations has been checked. As the participants of this study were randomly assigned to the conditions of the independent variable, this assumption has not been violated.

Second, the assumption regarding the linearity between the dependent variable and each of the independent variables has been checked. By visually assessing the partial regression plots, an approximately linear relationship has been found between the dependent variable and each of the covariates and the independent variable. Thus, the assumption has not been violated.

Third, the assumption of homoscedasticity has been checked by visually assessing a Scatterplot of studentized residuals against unstandardized predicted values. As the residuals did not have a specific pattern and were randomly scattered, the assumption has not been violated.

Fourth, the assumption of multicollinearity has been checked. Multicollinearity can be assessed by looking at the Variance Inflation Factors (VIF). Values that are greater than 4 show moderate multicollinearity, while values higher than 10 show strong multicollinearity issues (Malhotra, 2010). Therefore, all of the VIF scores of the independent variables were lower than 4 (see Table 4.2), which shows no multicollinearity issues and thus, this assumption has not been violated. Furthermore, the assumption of normality has been checked by visually assessing the P-P P-Plot. As the residuals were approximately normally distributed, the assumption has not been violated and the analysis can proceed.

(29)

28 not violated. Fourth, the assumption of multicollinearity has been checked. All of the VIF scores of the independent variables were lower than 4 (see Table 4.3 and Table 4.4), which shows no multicollinearity issues. Therefore, the assumption has not been violated. Lastly, the assumption of normality has been checked and as the residuals were approximately normally distributed, the assumption has not been violated.

Table 3.4: Estimated regression models

For Hypothesis 6, a Hayes process mediation analysis (regression) was performed and model number 4 was used to examine whether source credibility mediated the effect of social media marketing on purchase intentions. Covariates attitude toward the post and attitude toward the brand were also taken into account. When a bootstrapping method is being used, it is allowed to go further even if the data is not normally distributed (Carpenter and Bithell, 2000)

Model 1 (H3): PI = β0 +β1SC +

ε

PI = Purchase Intentions

SMM = Social Media Marketing

(dummy coded: 0 = company profile, 1 = influencer profile)

SC = Source Credibility BT = Brand Type

(30)

29

4. Results

4.1 Descriptive statistics and control variables

In this section, descriptive statistics of the demographics such as age, gender, and education of the sample will be described. A sample of 170 respondents was analyzed. It resulted that a higher percentage of the respondents were females (51.2%), rather than males (48.8%). The average age of the respondents was 30.8 years old, the youngest being 19 and the oldest being 66 years old. By gender, 51.2% of the participants were females, while 48.8% were males. Regarding their education, respondents had the following levels: 1.2% no education, 1.2% primary education, 29.4% high school diploma, 50% Bachelor`s degree, 15.3% Master`s degree and 2.9% had a Ph.D./Doctorate.

A linear regression analysis was conducted separately for each variable to determine which control variables to include in this study. The control variables were tested on the dependent variable, purchase intentions. Results have shown that age, gender, and education are not significant (p > .05) and therefore, these variables were not included in the analysis. On the other hand, attitude toward the post (p < .001) and attitude toward the brand (p < .001) were statistically significant when regressed on purchase intentions and were included in the study.

4.2 Correlation analysis

A Pearson correlation analysis has been conducted, which will provide further information about the relationship between two constructs and if a strong relationship is present (Malhotra, 2010). Therefore, the correlation between the independent variable (social media marketing), the dependent variable (purchase intentions), the mediator (source credibility), and the moderator (brand type) was analyzed.

(31)

30

Table 4.1: Pearson correlation results of the main variables Correlations Social media

marketing Purchase intentions Source credibility Brand type

Social media marketing - - - -

Purchase intentions .08 - - -

Source credibility .08 .65** - -

Brand type .00 .00 .16* -

Note: *Correlation is significant at the .05 level (two-tailed)

** Correlation is significant at the .01 level (two-tailed)

The results of the correlation analysis provided some first insights regarding the hypothesis that were previously developed. There was a strong correlation between source

credibility and purchase intentions (r = .64), and there was a small correlation between brand type and source credibility (r = .16).

4.3 Hypotheses testing

4.3.1 Hypothesis 1: Social media marketing and purchase intentions

A One-Way ANCOVA was conducted to determine the effect of social media marketing on purchase intentions after controlling for the attitude toward the brand and the attitude toward the post. Before performing the analysis, several assumptions have been checked. Therefore, there is not a statistically significant difference on the effect of social media marketing on purchase intentions between the two conditions, F(1,166) = 0.17, p = .67. Consequently, H1 is rejected.

4.3.2 Hypothesis 2: Social media marketing and source credibility

A One-Way ANCOVA was conducted to determine the effect of social media marketing on source credibility after controlling for the attitude toward the brand and the attitude toward the post. Before performing the analysis, the same assumptions as the ones used for hypothesis 1 have been checked. Therefore, there was not a statistically significant difference on the effect of social media marketing on source credibility between the two conditions, F(1,166) = 0.06, p = .80. Consequently, H2 is rejected.

4.3.3 Hypothesis 3: Source credibility and purchase intentions

(32)

31 brand and attitude toward the post. The covariates used in this analysis were included as independent variables. Before conducting the analysis, several assumptions have been checked.

The overall quality of the model was assessed by taking into consideration the model summary table. R-Square for the overall model was 51.7%, with an adjusted R-Square of 50.8%, which, according to scholars, is a substantial size effect (Cohen, 1988). Additionally, the overall statistical significance of the model was assessed using the ANOVA table. Therefore, the model is significant, F (3,166) = 59.21, p < .001.

Table 4.2: Multiple regression analysis with purchase intentions as the dependent variable Item Coefficients (β) P-value VIF

Variables

Constant - 1.89 < .001 -

Attitude toward the brand 0.15 .11 2.54

Attitude toward the post 0.43 < .001 2.84

Source credibility 0.50 < .001 1.94

R2 _.51 _- _-

Adjusted R2 .50 - -

F-value 59.21 < .001 -

The coefficients table (see Table 4.2) shows that the variables source credibility and attitude toward the post are significant with a p-value of .001, while the variable attitude toward the brand is not significant with a p-value of .118. Therefore, as source credibility positively affects purchase intentions, H3 is accepted.

4.3.4 Hypothesis 4: Brand type, social media marketing and purchase intentions

A Hayes` Process moderation analysis was conducted to determine if brand type (Z) acts as a moderator on the effect of social media marketing (X) on purchase intentions (Y). The effect of the covariates attitude toward the post and attitude toward the brand was also controlled. A number of 5000 bootstrapped samples with a confidence interval of 95% was used. The moderation model number 1 provided by Hayes (2003) was used as it can be seen in Figure 5.

(33)

32 Second, the interaction between social media marketing and brand type is not significant, p > 0.05. Additionally, the confidence interval was checked, and as 0 lies between the lower and the upper bound, the interaction is not significant. Although, the effect of the covariates attitude toward the post and attitude toward the brand was significant with a p-value of .001. Furthermore, the conditional effect of the moderation was checked and it was observed that in both conditions (non-luxury vs. luxury), the moderation effect is not significant. Consequently, H4 was rejected.

Table 4.3: Results of moderation analysis with purchase intention as the dependent variable Coefficients Std. Error p-value VIF

Variables

Constant -1.01 0.39 .01 -

Brand type - 0.29 0.31 .33 1.98

Social media marketing 0.02 0.31 .94 2.00

Interaction social media marketing – brand type

0.12 0.44 .78 2.86

Covariates

Attitude toward the post 0.64 0.12 < .001 2.48 Attitude toward the brand 0.29 0.11 < .001 2.60

R2 .46 - - -

F-value 27.69 - < .001 -

Figure 4.1: Results of moderation analysis with purchase intentions as the dependent variable

(34)

33

4.3.5 Hypothesis 5: Brand type, social media marketing and source credibility

A Hayes` Process moderation analysis was conducted to determine if brand type (Z) acts as a moderator on the effect of social media marketing (X) on source credibility (Y). The effect of the covariates attitude toward the post and attitude toward the brand was also controlled. A number of 5000 bootstrapped samples with a confidence interval of 95% was used. The moderation model number 1 provided by Hayes (2003) was used as it can be seen in Figure 4.1.

The overall significance of the model was checked and it was found significant, F(5,164) = 31.83, p < .001). The main effect of social media marketing on source credibility was statistically not significant with a β coefficient of - 0.01, p > .05 (see Table 4.4). As such, this result cannot be interpreted.

Second, the interaction between social media marketing and source credibility is not significant, p > 0.05. Additionally, the confidence interval was checked, and as 0 lies between the lower and the upper bound, the interaction is not significant. Furthermore, the conditional effect of the moderation was checked and it was observed that in both conditions (non-luxury vs. luxury), the moderation effect is not significant. Consequently, H5 was rejected.

Table 4.4: Results of moderation analysis with source credibility as the dependent variable Coefficients Std. Error p-value VIF

Variables

Constant 1.49 0.27 < .001 -

Brand type 0.19 0.31 .35 1.98

Social media marketing - 0.01 0.31 .96 2.00 Interaction social media

marketing – brand type

0.12 0.29 .69 2.86

Covariates

Attitude toward the post 0.47 0.08 < .001 2.48

Attitude toward the brand 0.18 0.07 .01 2.60

R2 .49 - - -

(35)

34

Figure 4.2: Results of moderation analysis with source credibility as the dependent variable

4.3.6 Hypothesis 6: Source credibility, social media marketing and purchase intentions

A Hayes` Process mediation analysis was conducted to determine if source credibility (M) acts as a mediator on the effect of social media marketing (X) on purchase intentions (Y). The effect of the covariates attitude toward the post and attitude toward the brand was also controlled. A number of 5000 bootstrapped samples with a confidence interval of 95% was used. The moderation model number 4 provided by Hayes (2003) was used as it can be seen in Figure 4.3.

The following paths have been analyzed as such: the a-path, which represents the effect of social media marketing (X) on source credibility (M) and the b-path, which represents the effect of source credibility on purchase intentions (Y). Afterward, the indirect effect, which is represented by the multiplication of a-path with b-path was assessed. Consequently, the last analyzed paths were the c-path (total effect), which represents the direct effect of social media marketing on purchase intentions and the c’-path (direct effect), which represents the effect of social media marketing on purchase intentions while controlling for the mediator source credibility (Hayes, 2009).

First, the effect of social media marketing on source credibility (a-path) has been analyzed. The effect of the covariates attitude toward the post and attitude toward the brand was also controlled. Therefore, results showed were significant, F(3,166) = 51.83, p < .001, but the effect of social media marketing on source credibility was not significant as social media marketing proved to have a low explanatory power of the model with β: 0.04, p > .05. On the other hand, the

(36)

35 covariates attitude toward the brand (β: 0.21, p < .001) and attitude toward the post (β: 0.45, p < .001) statistically significantly impacted source credibility.

Second, the effect of source credibility on purchase intentions (b-path) has been analyzed and the effect of the covariates was also taken into consideration. Therefore, results showed that the overall significance of the model is significant, F(4,165) = 44.21, p < .001. Also, the effect of source credibility was found to affect purchase intentions statistically significantly with β: 0.50, p- < .001. Regarding the covariates, attitude toward the brand was found not significant (p > .05), while attitude toward the post was found significant with β: 0.43, p < .001.

Third, by analyzing the indirect effect (the multiplication of a-path with b-path), the influence of social media marketing on purchase intentions was assessed. As 0 was found between the lower (BootLLCI = -.11) and the upper bound (BootULCI = .21) of the confidence interval, the indirect effect was not significant. Moreover, the standard error was 0.08, while the effect was .02. The total effect (c-path), which represents the mediation role of source credibility on the effect of social media marketing on purchase intentions, was analyzed further. Results were found to be statistically significant, F(3,166) = 45.98, p < .001. As 0 was found between the lower (BootLLCI = - .35) and the upper bound (BootULCI = .53) of the confidence interval, the total effect was not significant (p > .05). Moreover, the standard error was 0.22, while the effect was 0.09.

Figure 4.3: Results of mediation analysis with source credibility as mediator variable

(37)

36 Lastly, the direct effect (c’-path) of social media marketing on purchase intentions will be analyzed. As 0 was not found between the lower (BootLLCI = -.34) and the upper bound (BootULCI = .49) of the confidence interval, the total effect was not significant (p > .05). Moreover, the standard error was 0.21, while the effect was 0.07. According to the conditions of Baron and Kenny (1986), to have full mediation, the indirect effect between social media marketing and source credibility (a-path) and source credibility on purchase intentions (b-path) should be significant. Moreover, the direct effect should be non-significant. Therefore, in this case, there is no mediation as the only significant path is the b-path, the effect of source credibility on purchase intentions (p-value < .001).

4.4 Overview of the accepted/rejected hypotheses

In Table 4.5, the outcomes of the analyses are listed.

Table 4.5: Overview of accepted/rejected hypotheses

Hypotheses Accepted/rejected H1: For influencer profiles, the effect of social media marketing

on purchase intention is larger than for company profiles

Rejected

H2: For influencer profiles, the effect of social media marketing

on source credibility is larger than for company profiles.

Rejected

H3: Source credibility positively affects purchase intentions. Accepted

H4: For luxury brands, the effect of influencer profiles on

purchase intentions, compared to company profiles, is larger than for non-luxury brands.

Rejected

H5: For luxury brands, the effect of influencer profiles on source

credibility, compared to company profiles, is larger than for non-luxury brands.

Rejected

H6: Source credibility mediates the relationship between social

media marketing and purchase intentions

Rejected

(38)

37 results of the previous analyses were confirmed (see Table 4.5), in the sense that all of the hypotheses were rejected, except Hypothesis 3.

5. Discussion and conclusion

5.1 Summary

The results of this paper found no differences in the effect of social media marketing (company profiles versus influencer profiles) on purchase intentions. Similar results were obtained for the effect of social media marketing on source credibility and thus, there were no statistically significant differences between the conditions. On the other hand, the effect of source credibility was found to impact purchase intentions significantly. Furthermore, the analysis has shown that brand type does not have a moderating role for the effect of social media marketing on source credibility and neither for the effect of social media marketing on purchase intentions. Lastly, there was no mediating role for source credibility on the relationship between social media marketing and purchase intentions as only the b-path was found significant.

5.2 Discussion

In spite of Hypothesis 1, for influencer profiles, social media marketing did not have a larger effect on purchase intentions, compared to company profiles. More specifically, the effect of social media marketing on purchase intentions does not differ between consumers that were exposed to the influencer profile and those exposed to the company profile. Pornpitakpan (2004) and Lim et. al (2017) found that social media influencers have a positive impact on purchase intentions, but this paper found no evidence to support this. First, a possible explanation could be provided by the research of Djafarova and Rushworth (2017), who found that non-traditional influencers (bloggers and Instagram celebrities) have a higher impact on the consumer`s purchase intentions, compared to already established, traditional celebrities. In this study, a traditional celebrity, Cristiano Ronaldo, has been used and thus, could provide a possible explanation for the rejection of the hypothesis. Second, Lokithasan, Simon, Jasmin, & Othman (2019) found that gender may play a significant role in the buying decision as female consumers are more likely to acquire products from female influencers and male influencers have a higher impact on male consumers. This could be an explanation as in this study, only a male influencer has been used.

(39)

38 Bershka. The manipulation check revealed statistically significant differences, but also high scores for the non-luxury brand, which means that respondents might have perceived the brand Bershka as an above-average brand, regarding the perceived luxury, with a mean score of 4.21.

Hypothesis 2 was also not accepted, which means that for influencer profiles, social media

marketing did not have a larger effect on source credibility, compared to company profiles. Thus, this contradicts the findings of Veirman et al. (2017), Shuqair et al. (2016) and Schouten et al. (2019) who argue that online influencers are a more authentic and credible source as compared to social media marketing channels owned by companies. The reason for the rejection of this hypothesis might come from the research of Cooley and Parks-Yancy (2019), who found that even if online influencers have a positive impact on raising a product`s awareness, consumers have more trust in user-generated content, from people that they have met personally.

In line with Hypothesis 3 and contributing to existing literature, a significant positive relationship was found between source credibility and purchase intentions. This confirms the findings of Hui (2017), which found that there is a positive relationship between e-WOM, such as social media marketing and source credibility, when consumers put more emphasis on the credibility of the source rather than the type of social media that is being used. Besides, O’Cass & Lim (2001) and Wang & Yang (2010) found that the more credible a brand is, the more it affects purchase intentions. Therefore, the higher the credibility of a source, the higher the consumer`s purchase intentions.

Are influencers more credible advertisers of luxury brands?