The influence of disclosed sponsored versus non-disclosded Instagram posts on e-loyalty towards the influencer and brand: The interaction effect of brand credibility.

The influence of disclosed sponsored versus non-disclosded

Instagram posts on e-loyalty towards the influencer and brand:

The interaction effect of brand credibility.

Master’s Thesis Marketing Management

Author: Emilie Spaan

Adress: Pleiadenlaan 2-118, 9742NG Groningen E-mail: e.b.spaan@student.rug.nl

Phonenumber: +316 29487717 Student number: S3483320

Department: Faculty of Economics and Business Master: Marketing Management

2

Abstract

In today’s world, an increasing number of brands have started to collaborate with influencers on social media (i.e., Instagram) aiming to promote consumers’ loyalty towards the brand because loyalty can help them to gain an advantage over its competitors. This study investigated the extent to which sponsorship disclosure on Instagram posts is important to develop e-loyalty. More specifically, to what extent will such sponsorship disclosure influence the consumers’ e-loyalty towards the influencer and towards the brand, and to what extent does brand credibility strengthen this relationship? A 2 (sponsorship: disclosed vs. non-disclosed) x 2 (brand credibility: high vs. low) between-subject experimental design (N=187) showed that a disclosed Instagram post did not result in more e-loyalty towards the influencer (Hypothesis 1a) and brand (Hypothesis 1b) than a non-disclosed post. Furthermore, although high brand credibility resulted in more e-loyalty to the influencer and the brand (no specific hypotheses in the present research), there was no significant interaction effect of brand credibility on the effect of sponsorship on e-loyalty towards the influencer (Hypothesis 2) and towards the brand (Hypothesis 3). Additional analyses revealed that brand credibility has a stronger direct effect on e-loyalty towards the brand than on e-loyalty towards the influencer. These findings are discussed and in line with two important streams of research in the field of e-loyalty, that is, that is, sponsorship disclosure and brand credibility.

4

List of contents

Abstract ... 2 1. Introduction ... 6 2. Literature review ... 9 2.1 Loyalty ... 9

2.1.1 A specific type of loyalty: e-loyalty ... 10

2.2 Influence of social media on e-loyalty ... 11

2.2.1 Instagram as an important type of social media ... 11

2.2.2 Sponsorship disclosure versus non-disclosure in Instagram posts ... 12

2.3 Brand credibility ... 13

2.3.1 Brand credibility and e-loyalty ... 14

2.4 Conceptual framework ... 16

3. Method ... 17

3.1 Participants and sampling strategy ... 17

3.1.1 Excluding criteria ... 17

3.1.2 Sample size ... 17

3.2 Research Design ... 18

3.3 Pretests ... 19

3.3.1 Pretest 1: Choosing a neutral brand ... 20

3.3.2 Pretest 2: Choosing a (picture of a) neutral and realistic influencer ... 22

3.4 Materials ... 24

3.4.1 The Instagram timeline ... 24

3.4.2 Manipulations: and brand credibility ... 24

3.5 Procedure ... 25

3.6 Measures ... 26

3.6.1 Dependent variables: e-loyalty towards the brand and towards the influencer ... 26

3.6.2 Manipulation checks ... 26 3.7 Plan of analysis ... 27 3.7.2 Manipulations ... 27 3.7.2 Hypotheses ... 28 4. Results ... 31 4.1 Manipulation Check ... 31 4.1.1 Sponsorship disclosure ... 31 4.1.2 Brand credibility ... 31

4.2 Reliability and Validity ... 32

4.2.1 E-loyally towards the influencer ... 32

4.2.2 E-loyalty towards the brand ... 33

4.3 Hypothesis testing ... 33

4.3.1 The influence of a disclosed and a non-disclosed sponsorship on e-loyalty towards the influencer and e-loyalty towards the brand. ... 34

4.3.2 The influence of the interaction effect between sponsorship and brand credibility on e-loyalty towards the influencer ... 36

5

4.4 Additional analysis ... 39

4.4.1. The direct effect of brand credibility on e-loyalty towards the influencer and brand. ... 39

5. Discussion ... 41

5.1 Theoretical implications ... 43

5.2 Managerial implications ... 44

5.3 Limitations ... 44 5.4 Future research ... Fout! Bladwijzer niet gedefinieerd. Reference list ... Fout! Bladwijzer niet gedefinieerd.

6

1. Introduction

In today’s world, social media is becoming more prominent within the marketing field (Sokolova & Kefi, forthcoming). Social media can be seen as a community that a brand shares with its consumers. This community can increase brand loyalty more than traditional methods because communication via social media is more prone to be two-way communication (Zaglia, 2013). In contrast, communication via traditional media is more one-way communication, from a brand to the consumer (Alves, Fernandes & Raposo, 2016). Loyal consumers are the most important asset of a company because, whereas finding new consumers is very costly, retaining loyal consumers is more cost-efficient and easier (Lawfer, 2004).

Gensler, Völckner, Liu-Thompkins and Wiertz (2013) found that it is essential for a brand to use their social media properly because of the easy and open interaction on social media platforms. Of the different social media platforms, Instagram seems to be an excellent platform for a brand to connect with its consumers because 80% of the Instagram users do follow brands/organizations on Instagram (Van de Ketterij, 2019). Around 4.9 million people in The Netherlands have an Instagram account, and 2.7 million people scroll through their Instagram daily. In comparison to 2018, Instagram has grown with 20% (Oosterveer, 2019).

Instagram is the social media platform that experienced the most significant growth over the last year (Oosterveer, 2019). Furthermore, Instagram has the highest brand community and engagement of different social media platforms (Phua, Jin & Kim, 2017). Therefore, it is an excellent platform for a brand to invest (some) of their marketing budget into. For example, research suggests that it is vital for a brand to invest in Instagram for determining the right target audience for the brand (Djafarova & Rushworth, 2017).

7

between the recipient and the influencer to let the recipient follow up the influencers advice, and (3) the influencers must be able to influence behavioral intentions (Casaló, Flavián & Ibáñez-Sánchez, forthcoming). Leal, Hor-Meyll, and de Paula Pessôa (2014) found that an influencer at least needs one of the following three characteristics: (1) others should think the influencer has a good taste when it comes to their purchase decisions, (2) the influencer should be considered an expert in their domain, and (3) the influencer need to make substantial contributions to their domain and participate with high frequency. This means that when a brand wants to collaborate with an influencer, it is essential to pick an influencer who meets these criteria to get the most out of the collaboration.

When a brand decides to work together with influencers, it is essential that a brand select a social media influencer to whom its target group can ‘easily’ relate (Clarke, 2014). Furthermore, there must be a logical link between the influencer and the brand (Uzunoğlu & Kip, 2014). Previous research by Djafarova and Rushworth (2017) found that among young females’ influencers, like bloggers, ‘instafamous profiles,’ and YouTube personalities, have more influence on purchase decisions than ‘normal’-celebrities. Therefore, it is more interesting for brands to work together with influencers than with normal celebrities when it comes to purchasing intentions.

8

However, placing a disclosed sponsored Instagram post on the Instagram page of the influencer might also develop a certain extent of loyalty towards the influencer. So, who gets the loyalty of the ‘follower’? Will it go to L’Oréal Paris (brand), will it go to @nikkietutorials (influencer) or will it be to both or none of them? This research aims to examine the extent to which a disclosed versus non-disclosed sponsored Instagram post influences the e-loyalty towards the influencer and the e-e-loyalty towards the brand. We will also examine whether the credibility of the brand can further strengthen these relationships.

9

2. Literature review

2.1 Loyalty

The concept of loyalty has been studied for a long time, but there is no clear consensus about the definition of loyalty in the academic literature (Toufaily, Ricard & Perrien, 2013). Some researchers argue that for loyalty, two different conditions are needed (1) attitudinal preference, and (2) repeat purchase behavior (Dick & Basu, 1994). The attitudinal preference refers to which extent a brand/influencer is considered different from its competitors in the consumer’s mind, while purchase behavior refers to the extent to which people act on their loyalty.

Different researchers have used the four-stage model of Oliver (1999) as a basis to define and operationalize loyalty (e.g., McMullan & Gilmore, 2003). This model consists of four stages, (1) cognitive, (2) affective, (3) conative, and (4) action. This model is ‘special’ because it does not only look into the behavioral side of loyalty but also the attitudinal side. The lowest level of loyalty and the first stage within the model is the cognitive level. This level refers to functional parts like promotional and prices. The next stage is affective loyalty, and this stage refers to the attitudinal side of loyalty. When these first two stages are combined, these two will lead to preference. The third stage of loyalty is the conative stage, this stage expresses an intention and an urge to act (fourth stage) towards the company. For a brand or influencer, the ultimate goal should be to get to the stage that people act on their loyalty. Within the model of Oliver (1999), this relates to the fourth stage, action. However, this research will not measure actual purchase behavior; it is still the ultimate goal for brands/influencers within loyalty. Oliver (1999) states that loyalty is a multidimensional construct, in this research we will follow this view.

10

a more premium price and are more resistant to switch to competitors and, therefore, more likely to stay at the brand they are loyal to (Gommans, Krishnan & Sheffold, 2001). Both the brand and influencer start collaborating, among other things, to both gain financial benefits from the collaboration.

2.1.1 A specific type of loyalty: e-loyalty

11

influencer. This definition acknowledges the multi-dimensionality of e-loyalty and hereby its operationalization into seven sub-dimensions as proposed by Alfonzan and colleagues (forthcoming).

E-loyalty has often been conceptualized towards a brand or company. However, it can be conceptualized towards an influencer too. Within this research, we include both types of e-loyalty. E-loyalty towards the brand is about how much e-loyalty the consumer feels towards the brand featured in the Instagram post. While e-loyalty towards the influencers refers to how much e-loyalty the consumer feels towards the influencer featured in the Instagram post.

2.2 Influence of social media on e-loyalty

2.2.1 Instagram as an important type of social media

Social media is defined as “a group of Internet-based applications that build on the ideological and technological foundations of Web 2.0, and that allow the creation and exchange of User Generated Content” (Kaplan & Haenlein, 2010, p.61). A great advantage that social media has over traditional media (e.g., newspaper, television) is that the capacity of interaction between brand/influencer and consumers is greater (Phua, Jin & Kim, 2017). The growing capacity for interaction creates a context in which e-loyalty between the consumer and the influencer and/or brand can be developed. By using social media, a brand/influencer can communicate timely and direct to its end-consumer. Communicating through social media is a two-sided way of communication between a brand/influencer and its consumers. This means the consumer gains an active participating role (Kaplan & Haenlein, 2010). Findings from Laroche, Habibi & Richard (2012) found that because of this change in communication, social media platforms (e.g., Instagram) could help the brand to create more loyal consumers. Therefore, we focus on the online environment within this research.

12

Research has found that the number of followers that an Instagram page has affects the popularity of that account (De Veirman et al., 2017). These Instagram pages are considered as more likable, which might lead to more e-loyalty. Furthermore, Instagram is the most used social media platform when it comes to influencer marketing (Boerman, 2019). Therefore, Instagram will function as the research context for the present research.

2.2.2 Sponsorship disclosure versus non-disclosure in Instagram posts

For consumers, it has become harder to recognize sponsorship with the rise of the advertisement in the online environment (e.g., social media) than in traditional media (e.g., newspaper) (Stubb & Colliander, 2019). An Instagram post is considered sponsored when the influencer that posts the post is either getting financial compensation for it or when the influencer is getting the product/service central in the post for free (De Veirman & Hudders, 2019). Influencers are continuously sharing content about their lives and opinions about various topics, among which products, which makes it difficult for consumers to distinguish the sponsored posts form the non-sponsored posts (Gillin, 2008). To overcome this confusion on the consumer’s side and to create more transparency, advertisers and regulators have developed social media advertising regulations and policy standards to avoid that consumers might be misled by the ‘hidden’ sponsorship (FTC, 2017). The primary task of these regulations and policy standards is to make clear to a consumer that the message contains a form of sponsorship (Cain, 2011). When consumers notice that a post is disclosed, the process of ad recognition will start in the consumer’s mind (Boerman, Willemsen & Van der Aa, 2017). This process of ad recognition will make the consumer conscious about the fact that he/she is exposed to a message, containing a sponsorship instead of a ‘regular,’ non-sponsored message. To get more transparency and clarity into which post is a sponsored post on Instagram and which post is not, Instagram has created one universal disclosure (‘Paid partnership with [brand]’) (Boerman, 2019). This disclosure helps consumers to recognize paid partnerships that influencers might have with brands and make them aware of the sponsorship that this influencer is in with this brand.

13

their collaboration or not. Evans, Wojdynski, and Hoy (2018) found that a consumer can experience feelings of persuasion due to sponsorships of a brand/influencer. However, these feelings of persuasion can be mitigated by the transparency of the sponsorship. In addition, research has found the more transparent the sponsorship is, the fewer feelings of deception consumers will experience and accordingly reactance (Brehm 1966; Evans, Wojdynski, and Hoy 2018). Recent research has found that disclosure of sponsorships leads to more positive evaluation towards influencer and the brand since, with disclosure, the sponsorship becomes transparent (De Veirman & Hudders, 2019). Moreover, the more positive the brand evaluation is, the more e-loyalty consumers will have towards the brand and towards the influencer will be (Punniyamoorthy, Mahadevan, Shetty & Lakshmi, 2011). From these, we conclude that because e-loyalty includes both attitudinal and behavioral components, it can be assumed that therefore disclosing a sponsorship will also increase the consumers’ e-loyalty towards the influencers and towards the brand.

Therefore, we hypothesize:

H1a: A disclosed sponsored Instagram post will lead to more e-loyalty towards the influencer than a non-disclosed sponsored Instagram post.

H1b: A disclosed sponsored Instagram post will lead to more e-loyalty towards the brand than a non-disclosed sponsored Instagram post.\

2.3 Brand credibility

14

Other research shows that different factors can influence a brand’s credibility. Aaker (1991) found that the higher consumers perceive the quality of the brand, the lower the information costs are, and the less risk the consumers see in trusting the brand, the more credible they perceive the brand. In addition, credibility has found to be higher for brands that continuously ‘preach’ the same message (Erdem & Swait, 1998). From this we can conclude that it is beneficial for brands to be perceived as a high credible brand.

Johnson and Kaye (2015) found that while traditional media are seen as more credible than online media, social media gives consumers a form of gratification that traditional media is not able to provide. So, people turn to social media platforms, like Instagram, instead of traditional media for the need of satisfaction although they find social media less credible. Therefore, it might be harder for a brand to gain high brand credibility from its consumers in the online environment than it is in the offline environment.

2.3.1 Brand credibility and e-loyalty

A brand should aim to be perceived as a highly credible brand. Because consumers use brands as a signal to express information, and the more credible the consumer perceives the brand, the more likely the consumer chooses this brand (Erdem, Swait & Louviere, 2002). The credibility of a brand is found to be important to the brand, because the higher the credibility, the lower the economic decision-making cost and further it decreases the uncertainty that the consumer might feel towards the brand (Akdeniz, Calantone & Voorhees, 2012). Further, when the credibility of a brand drops, consumers might switch away from your brand to another brand that they rate as more credible (Johnson & Kaye, 2015). Moreover, when a consumer decides to switch away from your brand, you are losing a loyal consumer, which will decrease e-loyalty (López-Miguens & Vázquez, 2017).

15

(Erdem & Swait, 2004; Jeng, 2016). Furthermore, Erdem et al. (2002) found that the higher the brand credibility, the more positive the consumer perception of the brand will be (e.g., attitude). The strong relationship between brand credibility and attitudes, intentions, and behavior indicate that brand credibility is a vital factor for the development of e-loyalty towards the brand and e-loyalty towards the influencer.

Brand credibility is not only important because it has a direct effect on e-loyalty (e.g., Sweeney & Swait, 2008; Alam, Arshad & Shabbir, 2012; Ghorban & Tahernejad, 2012). Nevertheless, the effect of sponsorship disclosure on e-loyalty towards the brand and towards the influencer depends on it. Kang and Hustvedt (2013) found that the degree of transparency of a company affects the consumer’s attitude towards the company just as the intention to purchase and the action purchase behavior. Following the findings Kang and Hustvedt (2013), there can be said that the more transparent a brand is, the higher the brand credibility will be perceived by its consumers.

16

Therefore, we hypothesize:

H2: A high brand credibility will lead to more e-loyalty towards the influencer when including a disclosed rather than a non-disclosed sponsored Instagram post (Hypo 2a), and low brand credibility will lead to less e-loyalty towards the influencer when including non-disclosed rather than a disclosed sponsored Instagram post (Hypo 2b).

H3: A high brand credibility will lead to more e-loyalty towards the brand when including a disclosed rather than a non-disclosed sponsored Instagram post (Hypo 3a), and low brand credibility will lead to less e-loyalty towards the brand when including non-disclosed rather than a disclosed sponsored Instagram post (Hypo 3b).

2.4 Conceptual framework

From the previous literature review the conceptual framework as presented in Figure 1 is the basis of this research.

Figure 1. Conceptual framework.

Sponsored Instagram post disclosed versus non-disclosed

E-loyalty towards Influencer

Brand Credibility High versus Low

17

3. Method

This study examined if disclosed sponsored Instagram posts have a more positive relationship on the e-loyalty towards the influencer and the e-loyalty towards the brand than non-disclosed sponsored posts. Furthermore, the study further examined if brand credibility can strengthen this effect.

3.1 Participants and sampling strategy

The target group of this research were the 4,9 million people in the Netherlands who use Instagram (Oosterveer, 2019). The sampling technique was convenience sampling (Malhotra, 2010). The data needed to conduct the research was gathered via an online experiment, and participation was entirely voluntarily. The experiment was distributed different social media channels, such as Instagram, LinkedIn, and Facebook. The potential participants were provided with an online link, which has directly directed them to the online experiment that was exploited via the software Qualtrics.

3.1.1 Excluding criteria

To conduct the experiment properly, one filter question was asked at the start of the experiment. The filter question asked whether or not the respondent is a user of the social media platform Instagram. If the respondent had answered the question with ‘no,’ the respondent was directly directed to the end of the questionnaire. Because no representative or valid results could be drawn from a respondent who does not use Instagram. After the respondent answered that they are familiar with the social media platform, they proceeded to the real experiment.

3.1.2 Sample size

18

participants per condition to be able to measure the group differences, as the goal of this research was to compare the different conditions to each other.

In total 318 participants have started the survey; of them 245 participants have completed the survey. Fifty-five participants were excluded from the analysis because they were not familiar with the social media platform Instagram, and three responses were deleted because they were perceived as outliers, which left us with an actual sample of 187 participants. According to Cohen (1998), this sample size was big enough to measure the group differences because all of the four conditions consisted of more than 30 participants.

The analyses included 187 participants. The sample consisted of 70.1% female and 29.9% male. The participant’s ages ranged from 16 years to 74 years (Mage=27.20, SD=11.12). The

majority of the participants (74.9%) have an HBO bachelor’s degree or higher. Further, the greatest part (58.8%) has a yearly income of less than €15.000. These statistics suggest that the sample is relatively young and well-educated. This sample was a representative sample for the 2.7 daily Instagram users in the Netherlands (Oosterveer, 2019) because most of the Instagram users are younger than 39 years (Van der Veer, Broekee & Hoekstra, 2019).

3.2 Research Design

19

Table 1. Experimental design.

Condition Brand Credibility Sponsorship n

1 Low Disclosed 43

2 Low Non-disclosed 44

3 High Disclosed 53

4 High Non-disclosed 47

The experiment started by showing the participant an Instagram post in which both an influencer and a brand were explicitly visible. At the start of the experiment, the participant was randomly assigned to one condition for both the disclosure condition and brand credibility condition. In the sponsored Instagram post condition, the participant was able to read in the caption of the post (‘Paid partnership with [brand]’), to make clear that it was a sponsored Instagram post. In the non-sponsored Instagram post condition (‘Paid partnership with [brand]’) was not included in the caption of the Instagram post. The participant was either assigned to the high brand credibility condition or the low brand credibility condition. In the high brand credibility condition, the Instagram post showed comments from followers implying, that the brand featured in this post was a highly credible brand. In the low brand credibility condition, the comments from followers in the featured post have implied that it was a low credible brand. See Appendix A for the experimental conditions.

3.3 Pretests

20

Both pretests were distributed via WhatsApp by sending an anonymous link which directed the participant to the pretest. Only Dutch people were contacted as potential respondents for the pretest because all questionnaires within this research were in Dutch and therefore, probably mostly filled in by native Dutch speakers.

3.3.1 Pretest 1: Choosing a neutral brand

The first pretest aimed to choose an appropriate brand to use in the main questionnaire. We took four criteria into account to decide which brands to include in the second pretest, and consequently, into the main study.

First, we decided to only focus on brands that were associated with food and drinks. Food and drinks imply low involvement products (Verbeke & Vackier, 2004). Brands related to food and drinks are low involvement because the purchase decision to buy these products is made through a “force of habit” rather than through active information processing (Holmes & Crocker, 1987). Furthermore, in line with being a low-involvement product, food and drinks do not require a significant investment to buy (Szmigin & Piacentini, 2018). This means that these brands are interesting for the whole population of interest and does not exclude participants based on brands. Therefore, the brands used in the pretest are brands that are rather cheap and, therefore, do not require a significant investment to buy.

Second, Instagram influencers only promote existing brands in real life. Although participants might already hold attitudes to an existing brand because they know this brand, this would not be the case with a fictitious brand. However, to mimic reality as closely as possible, we decided to use real brands instead of fictitious brands within our experiment. Since the measured effects will better apply to the real context in this case, in reality, people might already hold attitudes to a promoted brand by an Instagram influencer.

21

wanted to be sure that the changes are significant that the participants were familiar with this brand.

Fourth, we decided to select brands that were most likely to be known by the population of interest, because we wanted that our tests came as close to reality as possible. At the same time, we wanted the brands to elicit an ‘as neutral as possible attitude.’ We aimed for neutral attitudes towards the brand because participants will not hold strong attitudes, either negative or positive, already towards that brand that might influence the manipulation of brand credibility.

In line with Evans et al. (2017), a pretest was conducted with ten different brands (i.e., Coca-Cola, Van Delft Pepernoten, Innocent Drinks, Sultana, Ben & Jerry’s, Peijnenburg, Nutella, Eat Natural, Crystal Clear, and Snickers). This first pretest aimed to find the brand that was considered as relatively neutral for the population of interest, meaning a neutral rating and as little deviation as possible. We aimed for neutral attitudes towards the brand because when brand credibility will be manipulated would be similar from a neutral starting point.

22

3.3.2 Pretest 2: Choosing a (picture of a) neutral and realistic influencer

The second pretest was conducted to find out which (picture of an) influencer was considered most neutral but still a realistic influencer. Evans et al. (2017) tested five different photos of women with a Dunkin’ Donuts coffee, these photo’s where presented to the participants to test for influencer likability, and to elicit the most neutral photo to use this one in the main questionnaire.

First, rather than using five photos’ featuring only women in the pretest like Evans et al. (2017), we used ten photos, featuring both men and women. Including males as credible influencers next to females were regarded as more desirable because, in reality, both women and men are Instagram Influencers.

Second, the photos that were used in our pretest are existing photos of influencers that placed this photo with an Innocent Drinks actually on their own Instagram feed. We choose to use existing photo’s rather than own made photos to stay as close to reality as possible. In line with Evans et al. (2017), we also used the most neutral photo in the main questionnaire because be manipulations would be similar from a neutral starting point.

Third, our aim was to start our main questionnaire with an as neutral as possible Instagram photo. To account for biased attitudes that might occur when respondents already know the person featured in the photo, participants were also given the option to choose for the option “I know the person on the person in the photo above”. We wanted to exclude photos with influencers to whom respondents are already familiar because this might have an influencer on the e-loyalty towards the influencer.

23

Twenty-six people were included in the second pretest. The measurements, “I am attracted to the person in the photo above,” was taken on a seven-point Likert scale (“1= totally disagree” to “7=totally agree”), with one extra answer option ‘I do know the person in the photo above.’ The second measurement, “The person in the photo above could be a realistic Instagram influencer,” also taken on a seven-point Likert scale. The questionnaire of this pretest can be found in Appendix C.

The results of the first measurement indicate that the photo shown in Figure 2 was considered as the most neutral photo out of the ten photos presented to the participants, as this photo showed a completely neutral response (M=4.00; SD=1.79). Respondents also showed a relatively neutral attitude towards photo 3 (M=4.38; SD=1.98) and photo 7 (M=3.50 ;SD=1.75) however, these photo’s showed relatively more variation of the mean indicating a slightly more polarized view towards these photo’s than photo shown in Figure 2 (See Appendix C for the Means and Standard Deviations of all photo’s). Therefore, based on this analysis photo, 10 was the best photo to use in the main questionnaire.

The results of the second measurement measured the extent to which the participant believed the influencer could be an Instagram influencer “The person in the photo above could be a realistic Instagram influencer.” Again, this item was measured on a seven-point Likert scale ranging from “1= totally disagree” to “7=totally agree”. The person in photo 6 was considered as the most realistic influencer (M=6.69 SD=1.54). Further, photo 3 (M=6.54

24

SD=1.96), photo 8 (M=6.31; SD=2.21), and photo 10 (M=6.00; SD=2.11) were considered as realistic photos. However, these photos showed diverse variation of the mean. From this, there can be concluded that the person in photo 6 is considered the most realistic influencer.

These two analyses combined we decided to use photo 10 in the main questionnaire. Although the person in photo 6 was considered as the most realistic influencer, this served not the primary goal, a photo as neutral as possible, of this pretest. We decided to use photo 10 because it is considered the most neutral photo of all, and still, the person in this photo was considered to be a realistic Instagram influencer.

3.4 Materials

3.4.1 The Instagram timeline

The Instagram post used in the main questionnaire featured the brand and the photo that followed from the results of the two pretests. We wanted to create an Instagram post, including the final picture depicting the influencer with the Innocent Drink, that would be perceived as realistic by participants. To mimic the Instagram layout, the website zeoob.com/generate-Instagram-post is used.

Furthermore, a realistic Instagram post should include likes and comments. To account for the influence of the number of likes and comments on e-loyalty towards the brand and influencer, we included the same number of likes and comments across the different experimental conditions. Further, the four conditions displayed the same photo and the same layout (see Appendix A).

3.4.2 Manipulations: and brand credibility

Within this research, there were two manipulations, namely (1) disclosed versus non-disclosed sponsorship, and (2) brand credibility high versus low.

25

used the phrase ‘Paid partnership with [brand]’ because this is the universal disclosure as used on Instagram (Boerman, 2019). In the caption below the Instagram post, ‘Paid partnership with Innocent Drinks’ was shown together with a regular caption if the participant was assigned to the disclosed condition. If the participant was assigned to the non-disclosed condition, the same regular caption will be shown but without the ‘Paid partnership with Innocent Drinks’ part.

The brand credibility high versus low manipulation was placed in the comment section below the Instagram post. In the high brand credibility condition, the words trustworthy and expertise were shown in the comment section. According to Erdem and Swait (2004), ‘trustworthiness’ and ‘expertise’ are the two components of credibility; hence they may prime participants towards believing the influencer is credible. In the low brand credibility condition, the antonyms of the words ‘trustworthy’ and ‘expert’ were featured in the comment section below the post, which included ‘dishonest’ (Synonym, 2019a) and ‘not specialized’ (Synonym, 2019b). The Instagram post of each experimental condition is shown in Appendix A.

3.5 Procedure

The main questionnaire was created in Qualtrics and was in Dutch, because the survey was distributed among Dutch people. To overcome that the questions did not measure what they supposed to, forward and backward translating was used (Sun & PJU, 2009). Two people, with good understanding of both the English and Dutch language, translated the original measurements from English to Dutch. Afterwards two others, also with good understanding of both languages, translated the Dutch versions back to English. These versions where compared, and most precise translation was used in the main questionnaire. The questionnaire can be found in Appendix D.

26

To only get participants who are actually familiar with Instagram and therefore were able to complete the questionnaire a filter question will be asked. The filter question asked whether or not the respondent is a user of the social media platform Instagram. If the response was no, the respondent was directed to the end of the experiment. If yes, participants proceeded to the next part of the questionnaire.

If the participant answered that he/she is a user of Instagram, the participant was asked to imagine that he/she was scrolling through their Instagram feed and where exposed the Instagram post as shown. They were shown a fictitious Instagram post, altered to the two manipulations, with its comments (all conditions shown in Appendix A). After the respondent had a thorough look, forced by a timer, at the Instagram post, he/she was directed to the next part.

The next part of the experiment will contain the questions concerning the manipulation checks (see section 3.7), and measurements for the dependent variables e-loyalty towards the brand and e-loyalty towards the influencer will be exposed to the respondent. The final part includes socio-demographic questions.

3.6 Measures

3.6.1 Dependent variables: e-loyalty towards the brand and towards the influencer

Two times 22 items were used to measure the dependent variables, e-loyalty towards the brand (22 items), and the e-loyalty towards the influencer (22 items). The measures were taken from a validated scale by Alfonzan et al. (forthcoming), although they were adapted to fit the context of this study. The 22 items distinguish seven different sub-dimensions of e-loyalty: transactional intention, transactional behavior, preference, contextual behavior, participation behavior, participation intention, and contextual intention. All items were measured on a seven-point Likert scale (“1= totally disagree” to “7=totally agree”).

3.6.2 Manipulation checks

27

the two-construct expertise and trustworthiness, as these dimensions are the most important to measure brand credibility. Spry, Pappu, and Cornwell (2009), found a Chronbach’s Alpha of .93 for the seven items that address expertise and trustworthiness for measuring brand credibility. All items were measured on a seven-point Likert scale (“1= totally disagree” to “7=totally agree”).

Sponsorship Disclosement. We were not able to address a validated scale that will be suitable for the manipulation check of the disclosure, therefore we created a measurement ourselves. The measurement consisted of three items, namely “A partnership between [brand] and [Influencer] is probable”, “[Brand] and [influencer] do have a paid partnership together” and “A collaboration between [brand] and [influencer] is likely.” All items were measured on a seven-point Likert scale (“1= totally disagree” to “7=totally agree”).

3.7 Plan of analysis

The software that has been used to analyze the data from the questionnaire is SPSS, version 24.

3.7.2 Manipulations

To check if the manipulations worked as desired, we used independent sample t-tests. Sponsorship and brand credibility were entered as independent variables and the manipulation check questions were used as the dependent variables. Before we used an independent samples t-test, we checked if all assumptions are met.

Assumptions t-test

The first assumptions addressed the scale of measurements. For an independent sample t-test, all measurements should have been taken on an ordinal or continuous scale. This assumption was met because the dependent variable was measured on a seven-point Likert scale.

28

The third assumption states that the distribution of the sample should be normal. This third assumption also relates closely to the fourth assumption, which states that the sample should be reasonably large and that per dependent variable group, the distribution should be normal, a bell-shaped curve. Because the sample size was smaller than 2000 participants, this was tested via a Shapiro Wilk test. This assumption was violated because not all groups were normally distributed (p>0.05). Although this assumption was violated, we still proceeded with this analysis because the Central Limit Theorem states that when all conditions have n>30, we may assume normality of distribution.

The last assumption concerns the homogeneity of variance. This assumption is met when the standard deviations of samples are approximately equal to each other. This was tested with Levene’s test for equal variances. Sponsorship (1=disclosed, 2=non-disclosed) entered as the independent variable, and the manipulation check was entered as the dependent variable. Levene’s test for equality of variances revealed a p>.05 namely, p=.38, which means equal variances were assumed.

Further, Brand credibility (1=low, 2=high) was entered as the independent variable and the brand credibility manipulation as the dependent variable. Levene’s test for equality of variances revealed p>.05, namely, p=.55 which means equal variances are assumed. From this, we can conclude that this last assumption was met.

We met all the assumptions to perform an independent samples t-test for the two manipulation checks, sponsorship, and brand credibility.

3.7.2 Hypotheses

29

Assumptions MANOVA.

The first assumption states that the two dependent variables, e-loyalty towards the brand and e-loyalty towards the influencer, should have been measured on an interval or ratio level. This assumption was met because our dependent variables were measured on a seven-point Likert scale.

The second assumption states that the independent variables, sponsorship disclosure, and brand credibility should both consist of two or more categorical groups. This assumption was met because the sponsorship was either disclosed or non-disclosed, and brand credibility was either high or low.

Thirdly, there must have been independence of observations. This assumption is also met because the participants received just one out of the four conditions. Therefore, we can state that there was an independence of observations.

The fourth assumption concerned that the sample size should be of adequate size. As describes in section 3.1.2, we aimed for at least 30 participants per condition. We met this assumption because we collected more than 30 participants per condition.

Fifth, the dataset cannot contain any outliers, these are values which are not following the normal pattern within the dataset. To detect if we had any outliers, we made two boxplots, one for the dependent variable e-loyalty towards the influencer and one for the dependent variable e-loyalty towards the brand. For the dependent variable e-loyalty towards the brand, no outliers where detected. For the dependent variable e-loyalty towards the influencer, three outliers were detected. These outliers were removed from the dataset because they could have an influencer on the results of the study. After removing the three participants, we conducted two new boxplots, one for each dependent variable, in these ones no outliers where shown. Therefore, after removing these three participants, this assumption was met.

30

Seventh, there must be a linear relationship between the dependent variables for all different combinations of groups for the independent variables. This assumption was tested via a scatterplot, and the scatterplot revealed a linear line for all different groups.

The eight assumption concerns the homogeneity of variance-covariance matrices. This assumption was tested by using a Box M test of equality of covariances matrices. This test observed if covariances matrices across the different groups on the dependent variable are assumed equal, this was the case. So, this assumption was met.

The ninth, and final assumptions, states that there must be no multicollinearity between the dependent variables. This means that both dependent variables, e-loyalty toward the brand and e-loyalty towards the influencer, should not correlate too much to each other. They correlate too much if r=.90 (Tabachnick & Fidell 2012). Our dependent variables revealed an r=.71, which is lower than r=.90 and therefore we can say that the last assumption was met.

31

4. Results

4.1 Manipulation Check

In order to check if the two manipulations, sponsorship disclosure (disclosed vs. non-disclosed), and brand credibility (high vs. low) had the desired outcome, two manipulation checks were conducted.

4.1.1 Sponsorship disclosure

The sponsorship manipulation check consisted of 3 items, “A partnership between Innocent Drinks and Tom de Vries (the influencer) is likely,” “Innocent Drinks and Tom de Vries (the influencer) do have a paid partnership,” and “ A collaboration between Innocent Drinks and Tom de Vries (the influencer) is probable.” We tested the internal consistency between these three items, Cronbach’s alpha yielded .79, which is higher than of .60 which is the minimum to combine the items into one variable (Peterson, 2013). We computed a new variable ‘sponsorship’ by summing the three items and divided this by three (M=5.72, SD=1.26).

In order to analyze if there is a difference between the participants who received the disclosed versus non-disclosed sponsorship condition, we performed an independent samples t-test. As the dependent variable, we entered sponsorship, and as the independent variable, the conditions (1=disclosed, 2=non-disclosed) the participant was assigned to. The participants who received the disclosed sponsorship condition (M=5.80, SD=1.29) perceived a slightly higher degree of sponsorship, than the participants assigned to the non-disclosed sponsorship (M=5.61, SD=1.23) condition. Nevertheless, the independent samples t-test did not yield a significant outcome t(182)=-1.02, p=.31, which is not significant. Therefore, our manipulation of sponsorship was unsuccessful.

4.1.2 Brand credibility

32

one variable (Peterson, 2013). We computed a new variable ‘brand credibility’ by summing the seven items and divided this by seven (M=4.22, SD=1.18).

In order to analyze if there is a difference between the participants who received the high versus low brand credibility condition, we performed another independent samples t-test. As the dependent variable, we entered brand credibility, and as the independent variable, the conditions (1=low, 2=high) the participant was assigned to. The means indicate that the participants who received the low brand credibility condition (M=3.88, SD=1.19) perceived the brand as less credible than the participants in the high brand credibility condition (M=4.52, SD=1.10). The independent samples t-test revealed that these differences were significant, t(184)=-3,82, p<.001. Therefore, our manipulation of credibility was successful.

4.2 Reliability and Validity

For both dependent variables, the scale developed and validated by Alfonzan et al. (forthcoming) was used; this scale consists of 22 items per dependent variable. To reduce these items into fewer variables, we conducted a Principal Component Analysis (PCA). A PCA is used to reduce a large quantity dataset by finding a common variance by finding underlying dimensions or tot test if hypothesized dimensions also exist in the dataset (Malhotra, 2010). Further, for both dependent variables, a Cronbach’s alpha reliability test is conducted to see if we are allowed to merge different items into one new variable.

4.2.1 E-loyally towards the influencer

Another PCA was conducted for the 22 items intended to measure the e-loyalty towards the brand, also taken from the same scale by Alfonzan et al. (forthcoming). PCA revealed a KMO=.91, and Bartlett’s test of sphericity yielded a significant score of p<.001. The communalities were all >.40. Therefore, we proceeded with all 22 items.

33

a new variable ‘e-loyalty towards the influencer’. Which was computed by summing up the 22 items and then dividing by 22 (M=2.16, SD=0.87).

4.2.2 E-loyalty towards the brand

A PCA was conducted with all 22 items measuring the participant’s e-loyalty towards the brand. The Kaiser-Meyer-Olkin (KMO) test revealed that it was a sufficient sample to proceed with the PCA, KMO=.89. Further, Bartlett’s test of sphericity yielded a significant score, p<.001. The communalities all need to be >.40, although one item revealed a slightly lower score (.35), we proceeded with all 22 items because the influencer of this item will be neglectable on the overall score.

In addition to the PCA analysis, we conducted a Cronbach’s alpha reliability analysis on the 22 items intended to measure e-loyalty towards the brand. The Cronbach’s alpha revealed a ⍺=.94, which means all 22 items are highly correlated. Therefore, we were allowed to generate a new variable ‘e-loyalty towards the brand’, which was computed by summing up the 22 items and then dividing by 22 (M=2.64, SD=0.95).

4.3 Hypothesis testing

34

Table 2. Multivariate tests, Wilks Lambda.

Effect Value F Hypothesis df Error df p

Partial Eta Square d Intercept .10 718.53 2.00 167.00 .00 .90 Manipulation sponsorship disclosure .99 0.32 2.00 167.00 .73 .01 Manipulation brand credibility .96 4.06 2.00 167.00 .02 .05 Manipulation sponsorship disclosure * brand credibility .99 0.28 2.00 167.00 .76 .01

Table 3. Test of between-subject effects. Interaction effect of Sponsorship and Brand Credibility on E-loyalty towards the brand and E-loyalty towards the influencer, R=.47.

Independent Variable Dependent Variable df F p

Sponsorship (disclosed

vs. non-disclosed) Brand e-loyalty 1 7.67 .01

Influencer e-loyalty 1 5.81 .02

Brand Credibility (high

vs. low) Brand e-loyalty 1 0.01 .97

Influencer e-loyalty 1 0.37 .54

Sponsorship * Brand

Credibility Brand e-loyalty 2 0.57 .45

Influencer e-loyalty 2 0.29 .59

4.3.1 The influence of a disclosed and a non-disclosed sponsorship on e-loyalty towards the influencer and e-loyalty towards the brand.

35

To analyze whether or not the e-loyalty towards the influencer and the e-loyalty towards the brand are influenced by if the sponsorship was disclosed or not, we performed a two-way MANOVA. The mean scores of the non-disclosed versus disclosed sponsorship conditions (see Figure 3) showed that the participants who were exposed to the disclosed sponsorship condition (M=2.64, SD=1.07) showed slightly more e-loyalty towards the brand than the participants who were exposed to the non-disclosed sponsorship condition (M=2.60, SD=0.80). The small differences in mean scores between the two conditions implied that sponsorship disclosure did not influence the e-loyalty towards the brand. The means for non-disclosed versus disclosed sponsorship on e-loyalty towards the influencer, showed a similar pattern. That is, participants in the disclosed sponsorship condition (M=2.19, SD=.92) showed slightly more e-loyalty towards the influencer than participants in the non-disclosed sponsorship (M=2.09, SD=0.72). The multivariate analysis revealed that these differences in mean were non-significant, F(2, 167)=0.32, p=.73; Wilks’ Λ = .99 (see Table 2), hereby rejecting Hypothesis 1a and 1b.

Figure 3. Mean differences between sponsorship (disclosed vs non-disclosed) on e-loyalty towards the brand and influencer. Note: 95% confidence interval.

2,64 2,19 2,6 2,09 0 0,5 1 1,5 2 2,5 3

E-loyalty toward the brand E-loyalty toward the influencer

36

4.3.2 The influence of the interaction effect between sponsorship and brand credibility on e-loyalty towards the influencer

Hypothesis 2a predicted that high brand credibility will lead to more e-loyalty towards the influencer when including a disclosed rather than a non-disclosed sponsored Instagram post (Hypo 2a), and low brand credibility will lead to less e-loyalty towards the influencer when including non-disclosed rather than a disclosed sponsored Instagram post (Hypo 2b).

37

Figure 4. Interaction between sponsorship and brand credibility on e-loyalty towards the Influencer. Note: 95% confidence interval.

4.3.3 The influence of the interaction effect between sponsorship and brand credibility on e-loyalty towards the brand.

Hypothesis 3 predicted that high brand credibility will lead to more e-loyalty towards the brand when including a disclosed rather than a non-disclosed sponsored Instagram post (Hypo 3a), and low brand credibility will lead to less e-loyalty towards the brand when including non-disclosed rather than a disclosed sponsored Instagram post (Hypo 3b).

In order to analyze whether there was a significant interaction effect of brand credibility and sponsorship disclosure on e-loyalty towards the brand, we conducted a two-way MANOVA. The means scores of the interaction between a disclosed sponsorship and high brand credibility condition versus the disclosed and low brand credibility condition (see Figure 5) show that the participants who received the disclosed sponsorship and high brand credibility (M=2.85, SD=1.13) showed more e-loyalty towards the brand than participants who were exposed to the brand, than participants who received the disclosed sponsorship and low brand credibility condition (M=2.35, SD=0.94). The differences in mean imply no interaction effect between the two conditions of disclosed sponsorship and brand credibility, both high and low. The means of the interaction between non-disclosed sponsorship with high brand credibility versus low brand credibility followed the same

2,35 2,2 1,97 1,96 0 0,5 1 1,5 2 2,5 3

Disclosed sponsorshp Non-disclosed sponsorship

E-lo ya lty to w ar ds th e in flu en ce r

38

pattern (see Figure 5). That is, participants who received the non-disclosed sponsorship, and high brand credibility condition (M=2.74, SD=0.78) showed slightly more e-loyalty towards the brand than participants in the non-disclosed sponsorship, and low brand credibility condition (M=2.45, SD=0.81). The Test of between-subjects effect yielded a score of F(1,171)=0.57, p=.45 (see table 3). Therefore, Hypothesis 3 was neither supported.

Figure 5. Interaction between sponsorship and brand credibility on e-loyalty towards the brand. Note: 95% confidence interval.

The table below gives an overview of the results that are found while the hypotheses were tested (Table 4).

Table 4. Hypotheses overview.

Hypotheses Findings

H1a: A disclosed sponsored Instagram post will lead to more e-loyalty towards the influencer than a non-disclosed sponsored Instagram post.

Not supported

H1b: A disclosed sponsored Instagram post will lead to more e-loyalty towards the brand than a non-disclosed sponsored Instagram post. Not supported 2,85 _2,74 2,35 2,45 0 0,5 1 1,5 2 2,5 3

Disclosed sponsorshp Non-disclosed sponsorship

E-lo ya lty to w ar d t he b ra nd

39

H2: A high brand credibility will lead to more e-loyalty towards the

influencer when including a disclosed rather than a non-disclosed sponsored Instagram post (Hypo 2a), and low brand credibility will lead to less e-loyalty towards the influencer when including non-disclosed rather than a non-disclosed sponsored Instagram post (Hypo 2b).

Not supported

H3: A high brand credibility will lead to more e-loyalty towards the brand when including a disclosed rather than a non-disclosed sponsored Instagram post (Hypo 3a), and low brand credibility will lead to less e-loyalty towards the brand when including non-disclosed rather than a non-disclosed sponsored Instagram post (Hypo 3b).

Not supported

4.4 Additional analysis

4.4.1. The direct effect of brand credibility on e-loyalty towards the influencer and brand. The direct effect of brand credibility (high vs. low) on the two dependent variables, e-loyalty towards the brand and e-loyalty towards the influencer, are not taken into account in the conceptual framework as presented in section 2.5. Although after running the analyses providing answers to the hypotheses, we found some results that we would like to incorporate in our results.

40

the means for the high brand credibility versus low brand credibility condition on e-loyalty towards the brand show a larger difference, which implies a significant difference between these two conditions. That is, participants in the high brand credibility condition (M=2.80, SD=0.97) showed more e-loyalty towards the brand than participants in the low brand credibility condition (M=2.40; SD=0.87). The Multivariate Tests reveal a significant score of F(2, 167) = 4.06, p<.05; Wilks’ Λ=0.95 (see table 2). We can conclude that there is a significant difference between the e-loyalty towards the influencer and e-loyalty toward the brand based on whether brand credibility is high or low.

The Test of Between-Subjects effects showed that when brand credibility increases with, the e-loyalty towards the brand increases with 0.40. Whereas when the brand credibility increases with one, the e-loyalty towards the influencer will increases with 0.30. From which we can conclude that the effect of brand credibility (when increased with one) is on e-loyalty towards the brand is 0.10 (=0.40-0.30) stronger than the effect of brand credibility towards e-loyalty towards the influencer.

Figure 6. Mean differences brand credibility (high vs. low) on e-loyalty towards the brand and influencer. Note: 95% confidence interval.

2,28 2,8 1,96 2,4 0 0,5 1 1,5 2 2,5 3

E-loyalty toward the influencer E-loyalty toward the brand

41

5. Discussion

The present study examined the effects of two forms of a sponsored Instagram post (disclosed and non-disclosed) on the e-loyalty towards the brand and e-loyalty towards the influencer. Based on previous studies (Punniyamoorthy, Mahadevan, Shetty & Lakshmi, 2011; Veirman & Hudders, 2019), our main expectations were that a disclosed sponsored Instagram post would lead to more e-loyalty towards the influencer (H1a), as well as to a more e-loyalty towards the brand (H1b).

Hypotheses 1a and 1b proposed that a disclosed sponsored Instagram post would lead to more e-loyalty towards the influencer (H1a) and e-loyalty towards the brand (H1b) than a non-disclosed sponsorship. The results of the analyses on hypotheses 1a and 1b are insignificant. These findings are not in line with previous research (Punniyamoorthy, Mahadevan, Shetty & Lakshmi, 2011; Veirman & Hudders, 2019), which implies that there should be a significant relationship. The insignificance can be due to the fact that although a sponsorship should nowadays be disclosed by regulations of the FTC (2017), these disclosures are such a small part of transparency that the effect of the disclosure is, therefore too limited to yield a significant difference. Another explanation can be, that people already intent that an Instagram post is sponsored, even when this sponship is not disclosed. This is based on the fact that they see a brand name in the Instagram post and therefore associate this so strongly with the brand (Virutamasen, Wongpreedee & Kungnungwut, 2015), that they automatically perceive it as a sponsorship. In addition, an explanation for this insignificant result can be that the manipulation for sponsorship (disclosed vs. non-disclosed) did not work as desired, there was no significant difference shown between the participants who were exposed to a disclosed condition compared to a non-disclosed condition. As a result of the unsuccessful manipulation, it might be the case that participants who received the non-disclosed sponsorship condition still assumed that the sponsorship was disclosed due to the prominent appearance of the brand.

42

Arshad & Shabbir, 2012; Ghorban & Tahernejad, 2012) has already established that there is a direct effect of brand credibly towards loyalty. Complimentary to this finding, this research investigated if there is an interaction effect between sponsorship disclosure and brand credibility. Based on Kang and Hustvedt (2013), and the congruency-principle (Rokeach & Rothman, 1965) we expect that high brand credibility will lead to more e-loyalty towards the influencer when including a disclosed rather than a non-disclosed sponsored Instagram post (Hypo 2a), and low brand credibility will lead to less e-loyalty towards the influencer when including non-disclosed rather than a disclosed sponsored Instagram post (Hypo 2b). Further, we expect the same effects on e-loyalty towards the brand. Namely, that high brand credibility will lead to more e-loyalty towards the brand when including a disclosed rather than a non-disclosed sponsored Instagram post (Hypo 3a), and low brand credibility will lead to less e-loyalty towards the brand when including non-disclosed rather than a disclosed sponsored Instagram post (Hypo 3b).

The results of the analyses on hypotheses 2 and 3 yielded insignificant results. While Evans, Wojdynski, and Hoy (2018) argued that transparency could mitigate the feelings of persuasion that a consumer might hold towards a sponsorship. The insignificant result can be due to the fact that although sponsorship disclosure and brand credibility hold aspects of transparency. Nevertheless, even together, they do not create enough transparency to mitigate the feelings of persuasion by the consumer. For example, this might be because it is not exposed if the influencer gets either financial compensation for placing the Instagram post, or that the influencer gets the product featured in the Instagram post for free. Finally, the interference of the unsuccessful manipulation of sponsorship disclosure might lead to the non-significant outcome, and therefore people followed their thoughts rather than the manipulation.

43

direct effect of brand credibility on loyalty. Our research showed results in line with these previous researches. Namely, brand credibility has a significant direct effect on e-loyalty towards the influencer and e-loyalty towards the brand. Nevertheless, our research contributes because, like previous research, we found there is a significant direct effect between brand credibility and e-loyalty. Moreover, we extend this finding by showing that the effect of brand credibility is stronger towards the e-loyalty towards the brand than the e-loyalty towards the influencer. The effect towards the brand is possibly stronger towards the brand because brand credibility in itself is stronger linked to the brand than towards the influencer.

5.1 Theoretical implications

Our findings build on the existing research in the field of e-loyalty. Our research shows that whether or not a sponsored Instagram post is disclosed does not affect the e-loyalty towards the influencer and towards the brand. While Kang and Hustvedt (2013) found that transparency of sponsorship, disclosing a sponsorship, can mitigate the feeling of persuasion that a consumer can experience when being exposed to a sponsorship. We can contribute to these findings that disclosure, while it is a form of transparency, does not provide enough transparency to a consumer that it can mitigate the feelings of persuasion when being exposed to a sponsorship. In addition, also the combination of sponsorship disclosure and brand credibility together do not generate enough transparency, so the consumer does not feel persuaded by the sponsorship.

44

5.2 Managerial implications

This study provides useful insights for brands and influencers First, if a brand or influencer wants to increase the degree of e-loyalty from its consumers, it does not matter for both if they disclose the sponsorship or not. Second, a brand should aim to be perceived as a highly credible brand because the more the brand credibility of the brand is, the more e-loyalty the consumer will have towards the brand. Third, an influencer should aim to work with brands that are perceived as a highly credible brand by consumers. Because the higher the brand credibility of the brand with whom the influencer collaborates is perceived, the more e-loyalty the influencer will gain from its consumers/followers.

5.3 Limitations

This study has some limitations which will be discussed. First, the manipulation check of sponsorship showed that the manipulation did not work as desired. It might be the case that just the writing of a sponsorship disclosure underneath an Instagram post is a too small part of transparency, that it did not have a significant effect. Alternatively, that just the writing of a sponsorship disclosure underneath an Instagram post was not strong enough to stop brand associations, which automatically imply a sponsorship even if the participant was assigned to the non-disclosure condition. Furthermore, this might have affected the results of the hypothesis.

45

5.4 Future research

Future research should aim at providing more insights into the effects of sponsorship disclosure as regulated by the FTC (2017) because this research found that the sponsorship disclosure does not influence the e-loyalty towards the influencer and the brand. This means that a sponsorship disclosure does not provide enough transparency to mitigate the feelings of persuasion. It is interesting to look into if it may increase feelings of persuasion at a consumer because, via the sponsorship disclosure, the consumer is made explicitly aware of the sponsorship. While also without the disclosure, the consumer already thinks it is a sponsorship, also if the sponsorship is not disclosed due to the prominent brand name.

Another interesting proposal for future research is to look into if influencer credibility could also affect both e-loyalty towards the brand and e-loyalty towards the influencer. Furthermore, following the congruency principle, if the e-loyalty of the consumer becomes more if both brand credibility and influencer credibility will be high. And, if the e-loyalty of the consumer will be less when both brand credibility and influencer credibility are low.