Disclosure – Does it harm or benefit influencer marketing
on Instagram?
A study on the effectiveness of different influencer types on purchase intention
when mediated through source credibility and moderated by disclosure.
By Hannah F. C. Daller
2
Disclosure – Does it harm or benefit influencer marketing
on Instagram?
A study on the effectiveness of different influencer types on purchase intention
when mediated through source credibility and moderated by disclosure.
By Hannah F. C. Daller
S3845192
University of Groningen, Faculty of Economics and Business, M.Sc. Marketing
Master Thesis
13
thof January 2020
Antaresstraat 35, 9742 LA Groningen, Netherlands
0049 15734794167
h.f.c.daller@student.rug.nl
3
Abstract
Digital native advertising is commonly used on Instagram. Therefore, the line between editorial content and advertisement becomes unclear. As a result, disclosure of sponsorships increasingly receives attention from law and the media. For practitioners it is important to know about the impact of disclosure on their marketing campaigns and what kind of role different types of influencers play in that context. This study uses a 2x2 (no disclosure vs. disclosure, traditional vs. non-traditional influencer) between subject, experimental design with 197 respondents to investigate the effectiveness of traditional and non-traditional influencers on purchase intention. The mediation effect of source credibility and the moderation effect of disclosure on the relationship between the type of influencer and source credibility, as well as on the relationship between source credibility and purchase intention are also included in the study. The results of the analysis show that the type of influencer does not affect source credibility or purchase intention. Source credibility positively influences purchase intention, indicating that the more credible influencers are perceived, the higher the purchase intention of the Instagram user will be. However, source credibility does not explain the underlying mechanism between the type of influencer and purchase intention as a mediator. Disclosure does neither affect the relationship between the type of influencer and source credibility, nor the relationship between source credibility and purchase intention. Interestingly, the control variable attitude towards the influencer has a positive impact on source credibility. Also, the control variable product involvement is found to positively influence source credibility and purchase intention.
4
Preface
It makes me very proud to present the results of my work to you. Writing the thesis was fun and very interesting. Thanks to the broad topic of “branded digital content” I could get creative and conducted my research in the field of influencer marketing on Instagram. Since I am active on Instagram, it was exciting to combine my personal interests and knowledge with the project itself.
I want to thank my supervisor dr. Janny Hoekstra for spending so much time and effort on guiding me through the process of writing my master thesis. The enjoyable working environment she created during the meetings was very much appreciated. I was challenged to improve my work and could always count on valuable feedback, along with some laughs. A huge “thank you” goes to my family and boyfriend for supporting me during this last step of my academic education, even though we were 680km apart.
Hannah Daller
5
Table of Contents
1 Introduction ... 6
2 Theoretical Framework ... 8
2.1 Conceptual Model ... 8
2.2 Literature Review and Hypotheses ... 9
2.2.1 Type of Influencer ... 9 2.2.2 Source Credibility ... 10 2.2.3 Disclosure ... 12 3 Methodology ... 15 3.1 Research Design ... 15 3.2 Data Collection ... 17 3.3 Construct Measurement ... 18 3.3.1 Measurement Scales ... 18 3.3.2 Factor Analysis ... 20
3.3.3 Reliability Analysis and Correlation Matrix ... 20
3.4 Manipulation Check ... 22
3.5 Method of Analysis ... 23
4 Results ... 24
4.1 Control Variables ... 24
4.2 Hypotheses Testing ... 24
4.2.1 Type of Influencer and Purchase Intention ... 25
4.2.2 Type of Influencer and Source Credibility ... 26
4.2.3 Source Credibility and Purchase Intention ... 27
4.2.4 The Mediating Effect of Source Credibility ... 27
4.2.5 The Moderating Effect of Disclosure on the Relationship between the Type of Influencer and Source Credibility ... 29
4.2.6 The Moderating Effect of Disclosure on the Relationship between Source Credibility and Purchase Intention ... 31
4.2.7 Testing the Conceptual Model ... 31
4.3 Additional Analyses ... 33
5 Conclusion ... 35
5.1 Discussion ... 35
5.1.1 Design of the Study ... 35
5.1.2 Substantive Explanations ... 36
5.2 Management Implications ... 38
5.3 Limitations and Future Research ... 39
6
1 Introduction
Marketing is facing major changes caused by the recent technological developments. These create new possibilities to interact with the customer (Lamberton & Stephen, 2016). Therefore, online advertising is becoming one of the biggest marketing trends: 44,6% of the total advertising expenditures are expected to be spent on this medium in 2020 (Zenith Optimedia, 2018). Branded digital content as a part of online advertisement is defined as “any paid advertising that takes the specific form and appearance of editorial content from the publisher itself” (Wojdynski & Evans, 2016, p.157). Since income from traditional advertising is decreasing, digital native advertising becomes more important as an alternative method that offers new forms of income generation (Probst, Grosswiele, & Pfleger, 2013). Digital native advertising is often used on social media platforms when an influencer, a person that is linked with many other users, advertises products or services of a company (Goldenberg, Han, Lehmann, & Hong, 2009; Ki & Kim, 2019). In 2020, investments in influencer marketing are estimated to be around eight billion US $ and the number of social media users is increasing rapidly worldwide (Statista, 2018; Statista, 2019-a). Instagram forexample is placed as the most important platform for influencer marketing (Association of National Advertisers, 2018). Consequently, influencer marketing on Instagram is becoming relevant for marketers.
Disclosure is an important topic in digital native advertising. Since digital native advertising combines editorial contents with advertisement, it might be a challenge for some consumers to separate them (Choi, Bang, Wojdynski, Lee, & Keib, 2018). Therefore, disclosure increasingly receives attention from American law, which urges influencers to disclose sponsorship (Electronic Code of Federal Regulation, 2009; Federal Trade Commission, 2013). In Europe disclosure is a controversial topic as well. For example, Pamela Reif, a German influencer for fitness and fashion with around four million followers on Instagram, went to court for the non-disclosure of advertisement on her account. Reif argued that posts that are solely private should not be disclosed. The court judged that a strict separation of private and commercial posts, especially for a younger audience, is not possible and disclosure is needed (Bräutigam, 2019). Having the legal discussions in mind, it would help managers to know more about the consequences of disclosure on a post’s effectiveness: Does it harm or benefit influencer marketing?
7
that disclosure can reduce the uncertainty about a post’s commercial intent. Since the influencer intentionally reveals the sponsorship, he or she is perceived as more credible (Carr & Hayes, 2014). On the other hand, recent studies found that disclosure can also lead to negative outcomes (see, e.g., Boerman, Van Reijmersdal, & Neijens, 2015; Boerman, Willemsen, & Van Der Aa, 2017; Campbell, Mohr, & Verlegh, 2013). Especially the higher recognition of digital native advertisement that is connected to disclosure can lead to negative attitudinal and behavioral outcomes (Wojdynski & Evans, 2016). Evans, Phua, Lim and Jun (2017) identified that the clear wording “Paid Ad” leads to high recognition of the post as advertisement and therefore to decreased attitudes and intentions.
In addition to disclosure, the influencers themselves are central to an influencer marketing setting since they play an important role in their followers’ decision making (Ki & Kim, 2019). Influencers are defined as being able to create a “change in the belief, attitude, or behavior of a person” (Erchul & Raven, 1997, p.138). This means that impacting other users’ online activities, creating online engagement like WOM and effecting brand awareness as well as purchase intentions can be achieved (Hughes, Swaminathan, & Brooks, 2019; Hutter, Hautz, Dennhardt, & Füller, 2013; Trusov, Bodapati, & Bucklin, 2010). Therefore, celebrities are a promising tool in digital native advertising when used as influencers. The Source Credibility Theory, with credibility being a construct of trustworthiness, expertise and attractiveness, is commonly used to explain why celebrity endorsers are effective in advertisement (Ohanian, 1990). Consumers are more likely to be persuaded by a celebrity that is high in these dimensions (Erdogan, 1999). Social media itself brings forth a new kind of celebrity. Marwick (2015) suggested a separation between traditional influencers (athletes, supermodels or singers) and “Instafamous” personalities (non-traditional influencers) that reached public recognition through their social media self-branding.
8
The thesis is structured as follows: In chapter 2 the theoretical framework including the conceptual model, the hypotheses and the literature review is explained. Chapter 3 describes the methodology used and chapter 4 summarizes the results of the research. Finally, a conclusion containing a discussion, management implications, limitations and possible future research directions is presented in chapter 5.
2 Theoretical Framework
2.1 Conceptual Model
Figure 2.1 depicts the conceptual model of the thesis with the central variables influencer type, source credibility, purchase intention and disclosure.
Figure 2.1: Conceptual model
9
influencer who gained popularity through self-branding on social media (Khamis, Ang, & Welling, 2017; Lin, Bruning, & Swarna, 2018). Few research was conducted comparing both types, but the indications are that non-traditional influencers have a larger positive effect on purchase intention compared to traditional influencers (H1) (Djafarova & Rushworth, 2017; Schouten, Janssen, & Verspaget, 2019). Also, non-traditional influencers are perceived as more credible than traditional influencers (H2) (Djafarova & Rushworth, 2017; Schouten et al., 2019). Source credibility is predicted to have a positive effect on purchase intention (Bergkvist & Zhou, 2016) (H3). Additionally, it is expected that source credibility partially mediates the relationship between type of influencer and purchase intention (H4) (Amos, Holmes, & Strutton, 2008). The detailed reasoning for the hypotheses can be found in section 2.2.
The control variables age, gender, Instagram usage, product involvement, attitude towards the influencer and influencer familiarity are used since they could also impact the researched relationships.
2.2 Literature Review and Hypotheses
2.2.1 Type of Influencer
In this thesis a distinction is made between traditional and non-traditional influencers. Traditional influencers are famous because they made a career in a profession like acting, singing, sports or modeling (Lin et al., 2018). An example is Selena Gomez who first became famous as an actress in television series like “Hannah Montana” or “Wizards of Waverly Place” and later started a music career with her debut album “Kiss & Tell” (Biography, 2019). Traditional influencers are an effective marketing tool as their opinions have an impact on the behaviour of the audience (Bearden, Netemeyer, & Teel, 1989). Additionally, a positive response to products can be achieved by utilizing traditional influencers in comparison to non-celebrities since their image gets linked with the advertised brand (Atkin & Block, 1983). This increases advertisement effectiveness (Erdogan, 1999) in terms of sales (Elberse & Verleun, 2012), purchase intention (Friedman & Friedman, 1979) and brand attitudes (Petty, Cacioppo, & Schumann, 1983).
10
blog and YouTube (Baxter-Wright, 2017). Like traditional influencers, non-traditional influencers can have positive effects when advertising products of a company. Vloggers for example increase brand perceptions and purchase intentions for luxury brands (Lee & Watkins, 2016). Also, vloggers product recommendations are often followed (Chapple & Cownie, 2017). Current literature on the effectiveness of traditional versus non-traditional influencers showed the following insights. In general, brands appear more relevant when advertised by a non-traditional influencer compared to a non-traditional influencer (Ki & Kim, 2019). Interviews conducted by Djafarova and Rushworth (2017) indicate that for young female consumers, non-traditional influencers are more effective in generating purchase behaviors and brand attitudes. Schouten et al. (2019) found that a non-traditional influencer is more effective in terms of creating purchase intention for different product categories. Therefore, the direct effect of the traditional influencers compared to the non-traditional influencers on purchase intention is hypothesized as follows:
H1: Non-traditional influencers have a larger positive effect on purchase intention than traditional influencers.
2.2.2 Source Credibility
11
H2: Non-traditional influencers are perceived as more credible than traditional influencers.
The Source Credibility Model developed by Ohanian (1990) includes three dimensions, namely attractiveness, trustworthiness and expertise. According to Amos et al. (2008) these elements describe best how the source, in this case the influencer, affects purchase intention. Attractiveness includes the perceived likability, elegance and physical attractiveness of the source. Trustworthiness is specified through the endorsers objectivity and honesty when providing information. Last, expertise is evaluated by assessing the skills, knowledge and qualifications of the influencer (Ohanian, 1990).
For traditional influencers the Source Credibility Model from Ohanian (1990) is confirmed. The endorsers’ credibility has positive effects on purchase intentions (Pornpitakpan, 2004). Also, if a traditional influencer is seen as an expert, positive brand attitudes can be achieved (Eisend & Langner, 2010). La Ferle and Choi (2005) confirmed that traditional influencers, in their case an actor, can lead to high attitudes towards the advertisement and the brand as well as to increased purchase intentions. Also, non-traditional influencers are assessed as reliable sources and can therefore influence the purchase decision of customers (Djafarova & Rushworth, 2017). Vloggers for example inspire consumers in topics like lifestyle. Because customers evaluate them as credible and especially trustworthy in this area, they often buy what vloggers recommend (Chapple & Cownie, 2017).
Research indicated that people that are assessed as credible cause positive product and brand evaluation (Bergkvist & Zhou, 2016). Also, behavioral intentions are more likely caused by a highly credible source compared to a less credible one (Sternthal, Phillips, & Dholakia, 1978). Atkin and Block (1983) explained why celebrities are perceived as trustworthy. Consumers think that they truly like the products and brands they endorse since they link their name with it. Therefore, advertising the brand has more to it than a financial motivation which leads to increased trust towards the celebrity. Based on this, the third hypothesis predicts:
H3: Source credibility has a positive effect on purchase intention.
12
Sarel, 1991) and can function as a mediator (La Ferle & Choi, 2005). Therefore, hypothesis four is formulated as follows:
H4: The effect of the influencer type on purchase intention will be partially mediated by source credibility.
A partial mediation is expected because a direct effect of the type of influencer on purchase intention is predicted by H1.
2.2.3 Disclosure
Social media is an advertisement platform where it is hard to tell the difference between entertainment and commercial intents. Thus, customers may experience problems separating editorial and advertised contents (Choi et al., 2018). In general, marketers should make advertisement on social media identifiable by disclosing the sponsorship of a post (Federal Trade Commission, 2013). Additionally, this rule holds for celebrities. They are asked to acknowledge product or brand endorsements (Electronic Code of Federal Regulation, 2009). These rules make sense since social media users perceive celebrities as online peers and thus the persuasive attempt is less evident compared to a brand’s post (Jin & Phua, 2014; Schouten et al., 2019).
One can expect that the effect of the type of influencer on source credibility is influenced by the presence or non-presence of disclosure in a post. For traditional influencers an interaction effect with disclosure was found in literature. Using disclosure (vs. no disclosure) in a post causes the followers to identify the post as advertisement. Therefore, it becomes evident to them that the traditional influencer does not endorse the product because of honestly liking but for monetary compensation. As a result, distrust is evoked (Boerman et al.,2017). This is in line with Campbell & Kirmani (2000) who stated that influence agents are assessed as less sincere when their ulterior motives are more accessible. Since trustworthiness is a dimension of source credibility, this reasoning suggests that the effect of the type of influencer on source credibility (H2) is strengthened by disclosure. Thus, indicating that the traditional influencers are perceived as even less credible than the non-traditional influencers when disclosure (vs. no disclosure) is present in a post.
13
the reader’s uncertainty regarding the quality and character of the opinion leader” (Carr & Hayes, 2014, p.46). This means that the non-traditional influencer can be trusted since the commercial intent of the post is disclosed voluntarily. The second reason for the finding can be explained by a positive, preexisting relationship between the advertiser and the follower where trust was built beforehand (Carl, 2008; Chapple & Cownie, 2017). Again, this line of argumentation implies that the effect of the type of influencer on source credibility (H2) is strengthened by disclosure, indicating that the non-traditional influencers are perceived as even more credible than the traditional influencers when disclosure (vs. no disclosure) is used. Thus, the fifth hypothesis states:
H5: Disclosure enhances the effect of type of influencer on source credibility.
14
2013). The attitudes towards a brand as well as the purchase intentions are lower when disclosure is used in blogs (Van Reijmersdal, Fransen, Van Noort, Opree, Vandeberg, Reusch, Van Lieshout, & Boerman, 2016). To sum up, research showed that disclosure (vs. no disclosure) causes negative outcomes for advertisement effectiveness like purchase intention in various media.
Having this in mind, the moderation effect of disclosure can be best explained by the Persuasion Knowledge Model (PKM). Advertising confronts consumers with persuasive messages. The more a customer experiences such messages, the greater becomes the understanding of them and therefore the persuasion knowledge grows (Friestad & Wright, 1994). It then “enables them to recognize, analyze, interpret, evaluate, and remember persuasion attempts and to select and execute coping tactics believed to be effective and appropriate” (Friestad & Wright, 1994, p.3). Coping tactics are for example disengaging, ignoring or counterarguing (Friestad & Wright, 1994). Applying the PKM to digital native advertising implies that when potential customers recognize the persuasion attempt of the post by noticing the sponsorship disclosure - which means that their persuasion knowledge gets activated - they will behave defensively towards the advertisement. This indicates that in the case of disclosure (vs. no disclosure) it matters less whether the source is credible. Since, as soon as the persuasion attempt is revealed by the sponsorship disclosure to the follower, the activated persuasion knowledge provokes coping behaviors which cause the positive effect of source credibility on purchase intention to be weaker. Therefore, the sixth hypothesis states:
15
3 Methodology
3.1 Research Design
In this thesis a 2x2 between subject, experimental design was used. The manipulation of two variables, namely the type of influencer and disclosure, led to four conditions which are depicted in figures 3.1, 3.2., 3.3, and 3.4. Each subject was randomly assigned to one of the conditions.
16
Figure 3.3: Non-traditional influencer, Figure 3.4: Non-traditional influencer, non-disclosure condition disclosure condition
17
To prepare the posts for the disclosure manipulation the text beneath the original Instagram posts was changed to “This lemonade tastes so delicious”. Also, a drink and a lemon emoji were added and the fictitious brand “Frish” was mentioned. This is adequate to a common caption on Instagram since many posts have these certain characteristics of a short text description and the use of emojis and tags. The same number of “likes” was inserted for all posts and the comment section was deleted to rule them out as confounds. Hence, the caption was the same in all conditions. The difference in the disclosure condition was created by the usage of the hashtag #PaidAd at the bottom of the post since this wording and positioning was found as most recognizable by research (Evans et al., 2017; Wojdynski & Evans, 2016). Additionally, the heading “Paid partnership with Frish” was inserted in the disclosure condition since it is a common way to indicate sponsorship on Instagram (Gomez, n.d.; Sugg, n.d.).
The brand that was used in the advertisement is a fictitious soft drink brand called “Frish”. A can of “Frish” was inserted into the original Instagram posts by placing it on the railing next to the influencer. The decision to use a fictional brand was based on findings of Wood and Burkhalter (2014). They found that celebrity influencers are able to especially increase attention to unknown brands which includes “Frish”. The decision for a soft drink was made because the product category appeals to males and females likewise. In this context the control variable product involvement was measured.
3.2 Data Collection
18
with the opportunity to enter an E-Mail address to participate in the lottery. On the last page the subjects were informed about the manipulations in the survey and thanked for their participation.
In total 271 surveys were filled in. Unfortunately, 73 participants started the survey but did not finish it. Most of the participants that did not complete the survey stopped at the beginning, for example when being exposed to the influencer’s post (n=38). The 73 incomplete surveys were not included into the data analysis. In addition, another survey was excluded because of an unrealistic high Instagram usage per day of 1000 minutes which equals around 16h. Eventually, 197 completed survey were considered for the data analysis. The distribution of the subjects across the four conditions is shown in table 3.1.
Table 3.1: Distribution of the participants across the four conditions
Conditions
Traditional influencer, non-disclosure: n=47 Traditional influencer, disclosure: n=50
Non-traditional influencer, non-disclosure: n=53 Non-traditional influencer, disclosure: n=47
Respondents were between 15 and 58 years old with a mean of 25,12 years. Most Instagram users are 18 to 34 years old (Statista, 2019-b). This trend is reflected in the data set, since 95,4% of the participants belong to this age group. The majority of subjects were female (61,4 %, n = 121), 38,6% were male (n = 76). The Instagram usage in minutes per day was between 0 and 240 minutes with a mean of 43,37 minutes. Most respondents visited Instagram for 60 minutes (19,3%) or 30 minutes (17,3 %).
3.3 Construct Measurement
3.3.1 Measurement Scales
19
Table 3.2: Overview over construct measurement, factor loadings and Cronbach’s alpha
Construct and Source Items and Extreme Points Factor loading
Cron-bach’s α Purchase Intention
Adapted from Chandran and Morwitz (2005)
1. How likely are you to buy the soft drink “Frish”? Highly unlikely/ Highly likely
2. How probable is it that you will purchase the soft drink “Frish”? Highly improbable/ Highly probable
3. How certain is it that you will purchase the soft drink “Frish”? Highly uncertain/ Highly certain
4. What chance is there that you will buy the soft drink “Frish”? No chance at all/ Very good chance
,881 ,906 ,906 ,904 ,946 (4 items) Source Credibility
Adapted from Ohanian (1990) Expertise Dimension Trustworthiness Dimension Attractiveness Dimension
I consider the influencer in the Instagram post as: 1. Not an Expert/ Expert
2. Inexperienced/ Experienced 3. Unknowledgeable/ Knowledgeable 4. Unqualified/ Qualified 5. Unskilled/ Skilled 1. Undependable/ Dependable 2. Dishonest/ Honest 3. Unreliable/ Reliable 4. Insincere/ Sincere 5. Untrustworthy/ Trustworthy 1. Unattractive/ Attractive 2. Not Classy/ Classy 3. Ugly/ Beautiful 4. Plain/ Elegant 5. Not Sexy/ Sexy
,823 ,845 ,844 ,853 ,785 ,596 ,843 ,878 ,833 ,855 ,817 (*) ,840 ,848 ,795 ,869 (14 items) ,905 (5 items) ,897 (5 items) ,863 (4 items) Product Involvement Adapted from Chandrasekaran (2004)
To what extent do you (dis)agree with the following statements? 1. I am particularly interested in soft drinks.
2. Given my personal interests, soft drinks are not very relevant to me. (r)
3. Overall, I am quite involved when I am purchasing soft drinks for personal use.
Strongly disagree/ Strongly agree
,871 ,822 ,726
,764 (3 items)
Attitude towards the Influencer
Adapted from Muehling (1987)
I consider the influencer in the Instagram post as: 1. Bad/ Good 2. Negative/ Positive 3. Unfavorable/ Favorable ,822 ,877 ,863 ,928 (3 items) Influencer Familiarity
Adapted from Van Noort, Voorveld, and Van Reijmersdal (2012)
1. To what extent are you familiar with the influencer in the post? Completely unfamiliar/ Completely familiar
(/) (/)
20
3.3.2 Factor Analysis
The goal of the factor analysis is to decrease the number of variables by putting together interrelated variables into a factor. It is therefore tested if the items load on their respective constructs. When conducting a factor analysis, several criteria have to be taken into account. The Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy should be greater than 0,5 and Bartlett’s test of sphericity should be significant to indicate that the factor analysis is appropriate (Malhotra, 2010). In addition, all communalities should have a value of at least 0,4 and factor loadings should be above 0,5 (Gijsenberg, 2019).
To obtain the factor loading results shown in table 3.2 the following steps were taken. First, the item “Given my personal interests, soft drinks are not very relevant to me” was reverse coded for the analysis. Then the 25 items were used together in a first factor analysis whereby influencer familiarity was not included since it was measured with a one item scale. As the method of analysis, the principal components analysis and the VARIMAX rotation were used. The results showed a KMO of 0,857, a significant test of Bartlett’s test of sphericity (p= 0,000) and all communalities had a value above 0,4. The item “I consider the influencer in the Instagram post as: “Not Classy/ Classy” only had a factor loading (0,534) very close to the critical value of 0,5 and a cross loading (0,355) on the construct “attitude towards the influencer” and was therefore excluded from the analysis.
The second factor analysis with the remaining 24 items was conducted and 6 factors, namely purchase intention, source credibility expertise, source credibility trustworthiness, source credibility attractiveness, product involvement and attitude towards the influencer were identified. The KMO was found to be 0,855, Bartlett’s test of sphericity was significant (p=0,000), all communalities had values above 0,4 and all factor loadings above 0,5 (see table 3.2). In addition, the components showed Eigenvalues greater than 1 and the cumulative percentage of variance was 76,985% which is well above the critical value of 60% (Malhotra, 2010).
3.3.3 Reliability Analysis and Correlation Matrix
21
Therefore, the new variables purchase intention, source credibility, product involvement, and attitude towards the influencer were calculated by summing up the items per construct and dividing the sum through the number of items. A special case is the variable source credibility which was constructed by combining the factors source expertise, source trustworthiness and source attractiveness since these are the three dimensions of source credibility. The new variables were used in the further analysis.
Table 3.3 shows the correlation matrix of the variables, including the correlation coefficients, means and standard deviations.
Table 3.3: Correlation matrix
Mean S.D. Purchase Intention Source Credibility Product Involvement Attitude towards the Influencer Purchase Intention 2,319 1,317 1 Source Credibility 4,375 ,888 ,284a 1 Product Involvement 3,289 1,422 ,297a ,122c 1 Attitude towards the Influencer 4,499 1,206 ,235a ,558a -,003 1
22
3.4 Manipulation Check
For the type of influencer as well as for the disclosure manipulation a pre-test was conducted to check if the manipulations work sufficiently. The 40 participants for the pre-test were approached at the Zernike Campus of the Rijksuniversiteit Groningen. Each of the four Instagram post conditions (see section 3.1) was used ten times. In addition, a short description about how the respective influencer got famous, based on section 2.2.1, was given. The respondents were advised to read through the description carefully and to have a good look at the post. Afterwards, they answered the two manipulation check questions depicted in table 3.4. The questions were measured on a 7-point scale. More specifically, the influencer manipulation was assessed on a semantic differential scale and the disclosure manipulation on a Likert scale.
Table 3.4: Overview over manipulation check questions
Manipulation Check Question with Extreme Points Influencer
Own scale based on Khamis et al. (2017); Lin et al. (2018)
The influencer I saw in the post became famous through: Being an athlete, singer, actor or model /
Self-branding on social media, for example as fashion, lifestyle, food or fitness blogger and/or YouTuber (semantic differential scale)
Disclosure
Own scale based on Wojdynski and Evans (2016)
To what extent do you (dis)agree with the following statement? The post I just saw was clear about its sponsorship.
Strongly disagree/ Strongly agree (Likert scale)
Two independent samples t-tests were performed to find out if the manipulations were successful. On the influencer scale, the traditional influencer (M=1,500, S.D.=0,688, n=20) was significantly different (p=0,000) in the right direction from the non-traditional influencer (M=6,600, S.D.=0,940, n=20). Also, on the disclosure scale the non-disclosure condition (M=5,150, S.D.=1,927, n=20) was significant different (p = 0,025) in the right direction from the disclosure condition (M=6,350, S.D.=1,226, n=20). This indicated that both manipulations were successful.
23
traditional influencer (M= 6,530, S.D.=1,159, n=100). On the disclosure scale, non-disclosure (M= 3,830, S.D.=2,188, n=100) was significantly different (p = 0,000) in the right direction from disclosure (M= 5,540, S.D.=1,714, n=97). Therefore, it can be concluded that both manipulations worked.
3.5 Method of Analysis
Testing the hypotheses was built up in two steps. First, all hypotheses were tested separately from each other. For H1, H2 and H5 an analysis of covariance (ANCOVA) was performed since the independent variables are binary and additional insight into the means and standard deviations per category are provided (Malhotra, 2010). For the hypotheses including continuous independent variables, linear regression was used. Therefore, H3 was analysed with a simple linear regression and H6 with a linear regression with interaction term. The mediation effect of H4 was tested with four linear regressions based on Baron and Kenny (1986) and also with the Hayes PROCESS model 4 (Hayes, 2017). The estimated models per hypothesis can be found in table 3.5. Two control variables were added to the models (see section 4.1).
In a second step, all hypotheses were tested together to investigate if the findings from the separate analysis hold when the whole theory is taken into account, as it is depicted in the conceptual model (see figure 2.1). This was done by using the Hayes PROCESS model 58 (Hayes, 2017).
Table 3.5: Estimated models for each hypothesis
Model 1 (H1): PI = β0 +β1ToI + β2Pr + β3AtI + ε PI = Purchase Intention
ToI =Type of Influencer
(dummy coded: 0 = traditional influencer, 1 = non-traditional influencer) SC = Source Credibility D = Disclosure (dummy coded: 0 = no disclosure, 1 = disclosure) Pr = Product Involvement AtI = Attitude towards the
Influencer
ε = Error Term Model 2 (H2): SC = β0 + β1ToI + β2Pr + β3AtI + ε
Model 3 (H3): PI = β0 +β1SC + β2Pr + β3AtI + ε Model 4 (H4): (1) PI = β0 +β1ToI + β2Pr + β3AtI + ε (2) SC = β0 +β1ToI + β2Pr + β3AtI + ε (3) PI = β0 +β1SC + β2Pr + β3AtI + ε (4) PI = β0 +β1ToI + β2SC + β3Pr + β4AtI + ε Model 5 (H5):
SC = β0 + β1ToI + β2D + β3ToI*D + β4Pr + β5AtI + ε
Model 6 (H6):
24
4 Results
4.1 Control Variables
To test which control variables should be included into the hypotheses testing, a linear regression analysis was conducted separately for each control variable. Therefore, the respective control variables were used as the independent variable and purchase intention as the dependent variable since it is the concluding dependent variable in the conceptual model (see figure 2.1). Instagram usage (p=0,758), gender (p=0,681), age (p=0,588) and influencer familiarity (p=0,258) were not found to be significant. Product involvement (p=0,000) and attitude towards the influencer (p=0,001) showed a significant outcome. When all control variables were tested together in one regression, the same variables were significant.
Summing up, it can be said that product involvement and attitude towards the influencer were significant on purchase intention and thus were included into all models of the following analysis as covariates.
4.2 Hypotheses Testing
First, the assumption of normal distribution of the data was investigated with the Shapiro-Wilk test which should be used for smaller sample sizes (Le Boedec, 2016). Only source credibility was found to be normally distributed (p=0,639), all other variables did not meet the assumption (p=0,000). Therefore, a nonparametric test would be suited to test the hypotheses. However, in the following analysis the data is treated as normally distributed.
25
4.2.1 Type of Influencer and Purchase Intention
H1 predicts that non-traditional influencers have a larger positive effect on purchase intention than traditional influencers. To test this hypothesis an ANCOVA was conducted. The descriptive statistics of the analysis can be found in table 4.1 and the results in table 4.2.
Table 4.1: Descriptive statistics with purchase intention as dependent variable (model 1)
Mean S.D. Number
Traditional Influencer 2,469 1,341 97
Non-traditional Influencer 2,173 1,283 100
Table 4.2: Results with purchase intention as dependent variable (model 1)
Model 1 df S.S. M.S. F p ηp2
Main Variable
Type of Influencer 1 1,550 1,550 1,034 ,310 ,005
Control Variables
Product Involvement 1 30,412 30,412 20,287 ,000 ,095 Attitude towards the Influencer 1 15,677 15,677 10,457 ,001 ,051
Error 193 289,330 1,499 Note: R2 = 0,149 (Adjusted R2 =0,135)
The type of influencer was not found to be significant (F(1,193)=1,034, p=0,310, ηp2 =0,005,
26
4.2.2 Type of Influencer and Source Credibility
The prediction of H2 is that non-traditional influencers are perceived as more credible than traditional influencers. An ANCOVA was carried out. The descriptive statistics of the analysis can be found in table 4.3 and the results in table 4.4.
Table 4.3: Descriptive statistics with source credibility as dependent variable (model 2)
Mean S.D. Number
Traditional Influencer 4,492 ,917 97
Non-traditional Influencer 4,262 ,848 100
Table 4.4: Results with source credibility as dependent variable (model 2)
Model 2 df S.S. M.S. F p ηp2
Main Variable
Type of Influencer 1 ,005 ,005 ,009 ,923 ,000
Control Variables
Product Involvement 1 2,382 2,382 4,424 ,037 ,022 Attitude towards the Influencer 1 45,515 45,515 84,511 ,000 ,305
Error 193 103,944 ,539 Note: R2 = 0,327 (Adjusted R2 = 0,317)
H2 is rejected because the type of influencer was not significant ((F(1,193)=0,009, p=0,923, ηp2=0,000, B=0,010, t=0,097). Nonetheless, the control variables product involvement
((F(1,193)=4,424, p=0,037, ηp2=0,022, B=0,078, t=2,103) and attitude towards the influencer
27
4.2.3 Source Credibility and Purchase Intention
H3 predicts that source credibility has a positive effect on purchase intention. To test the hypothesis, a linear regression was conducted. The results of model 3 can be found in table 4.5. The model was found to be significant (R2=0,164 ,F(3,193)=12,619, p=0,000). Likewise, source credibility had a positive, significant effect on purchase intention (B=,255, t=2,146, p=0,033). Therefore, H3 is supported, indicating that the more credible an influencer is perceived, the higher the purchase intention will be. In addition, the control variable product involvement had a positive, significant effect (B=,256, t=4,152, p=0,000) and the control variable attitude towards the influencer had a marginally significant, positive effect (B=,153, t=1,759, p=0,080) on purchase intention.
4.2.4 The Mediating Effect of Source Credibility
In order to test for the partial mediation effect of source credibility as predicted in H4, four linear regressions based on Baron and Kenny (1986) were modeled (see table 3.5, Model 4). Certain assumptions must be met to establish mediation. First, the effect of the type of influencer on purchase intention has to be significant (path c). Second, the type of influencer should effect source credibility significantly (path a). Third, source credibility has to have a significant effect on purchase intention (path b). And fourth, when it is controlled for the mediator source credibility, the effect of the type of influencer on purchase intention (path c´) has to be diminished for partial mediation or insignificant for full mediation (Baron & Kenny, 1986).
The results of the linear regression analysis can be found in table 4.6. All models were significant (p=0,000). Since the type of influencer did not have a significant effect on purchase intention (B=-,182, t=-1,017, p=0,310), the type of influencer did not have a significant effect on source credibility (B=-,010, t=-,097, p=0,923) and the type of influencer did not have a
diminished, significant effect on purchase intention when source credibility was controlled for
Table 4.5: Results with purchase intention as dependent variable
(model 3; standardized coefficients)
Model 3
Main Variable
Source Credibility ,172b
Control Variables
Product Involvement ,276a Attitude towards the Influencer ,140c R2 (Adjusted R2) ,164 (,151) R2 change ,164 F-value 12,619a
Note: a p-value < ,01; b p-value < ,05;
28
(B=-,180, t=-1,011, p=0,313), partial mediation cannot be assumed and H4 is rejected. Only the effect of source credibility on purchase intention was found to be significant (B=,255, t=2,146, p=0,033).
Table 4.6: Linear regression results (model 4; standardized coefficients)
Model 4 (1) ToI → PI (2) ToI → SC (3) SC → PI (4) ToI + SC → PI Main Variables Type of Influencer Source Credibility -,069 -,006 ,172b -,068 ,172b Control Variables Product Involvement ,299a ,124b ,276a ,278a
Attitude towards the Influencer ,221b ,557a ,140c ,125 R2 (Adjusted R2) ,149 (,135) ,327 (,317) ,164 (,151) ,168 (,151) R2 change ,149 ,327 ,164 ,168 F-value 11,228a 31,276a 12,619a 9,721a
Note: a p-value < ,01; b p-value < ,05; c p-value < ,10
The same four regressions were performed without the control variables. This time, the type of influencer was found to be marginally significant on source credibility (B=-,230, t=-1,827, p=0,069) and source credibility was significant on purchase intention (B=,422, t=4,139, p=0,000). However, H4 would still be rejected because the model of the effect of the influencer type on purchase intention was insignificant (R2=0,013, F(1,195)=2,517, p=0,114) and the type of influencer did not have a diminished, significant effect on purchase intention when source credibility was controlled for (B=-,203, t=-1,117, p=0,265).
A second analysis was conducted to test for the mediation effect of source credibility. To correct for the small sample size, a bootstrapping analysis with Hayes PROCESS model 4 was performed. Here, 0 should not be included in the confidence interval of the bootstrapped indirect effect to establish mediation (Hayes, 2017). Also, the assumptions of Baron and Kenny (1986) must hold.
29
variables product involvement and attitude towards the influencer were added. The models were found to be significant (p=0,000). Table 4.7 below shows the results of the mediation. Path b, the effect of source credibility on purchase intention showed significant results (B=0,255, t= 2,139, p=0,034). Since path a, c and c´ failed to show significance, the predicted partial mediation of H4 cannot be confirmed and H4 is rejected. In addition, the bootstrapped indirect effect from type of influencer on purchase intention through source credibility contained 0 in the confidence interval (95%CI [-0,059; 0,061]).
Table 4.7: Results of mediation
Path B S.E. t p LLCI ULCI
a -,010 ,107 -,097 ,923 -,222 ,202
b ,255 ,119 2,139 ,034 ,020 ,489
c -,182 ,179 -1,017 ,311 -,536 ,171
c´ -,180 ,178 -1,011 ,313 -,530 ,171
Note: a p-value < ,01; b p-value < ,05; c p-value < ,10
Repeating the analysis without the control variables as well leads to the rejection of H4. Path a (B=-0,230, t=-1,827, p=0,069) and b (B=0,407, t=3,962, p=0,000) indicated (marginally) significant effects but path c (B=-0,297, t=-1,587, p=0,114), path c’ (B=-0,203, t=-1,118, p=0,265) and the bootstrapped indirect effect (95%CI [-0,210; 0,008]) failed to show significance. Therefore, no partial mediation was found.
In summary, the findings of the four linear regressions and the findings of the Hayes PROCESS model 4 are in line and H4 is rejected.
4.2.5 The Moderating Effect of Disclosure on the Relationship between the Type of Influencer and Source Credibility
30
scores were below 10 (Hair et al., 2014), thus mean centering was not necessary. An ANCOVA was carried out, the descriptive statistics of the analysis can be found in table 4.8 and the results in table 4.9.
Table 4.8: Descriptive statistics with source credibility as dependent variable (model 5)
Mean S.D. Number
Traditional Influencer No Disclosure 4,465 ,833 47
Disclosure 4,517 ,997 50
Non-traditional Influencer No Disclosure 4,116 ,716 53
Disclosure 4,427 ,957 47
Table 4.9: Results with source credibility as dependent variable (model 5)
Model 5 df S.S. M.S. F p ηp2
Main Variable
Type of Influencer 1 ,003 ,003 ,005 ,942 ,000
Control Variables
Product Involvement 1 2,536 2,536 4,700 ,031 ,024 Attitude towards the Influencer 1 43,608 43,608 80,822 ,000 ,279
Moderator Disclosure 1 ,568 ,568 1,052 ,306 ,005 Interaction Effect Type of Influencer*Disclosure 1 ,311 ,311 ,576 ,449 ,003 Error 191 103,056 ,540 Note: R2 = 0,333 (Adjusted R2 = 0,315)
H5 is rejected because the interaction effect was not significant ((F(1,191)=0,576, p=0,449, ηp2=0,003) and therefore no moderation effect can be assumed. The control variables product
31
4.2.6 The Moderating Effect of Disclosure on the Relationship between Source Credibility and Purchase Intention
The prediction of H6 is that disclosure mitigates the effect of source credibility on purchase intention. To test this, a linear regression with an interaction term was conducted. Since the VIF scores of model 6 were above 10 (Hair et al., 2014), the independent variables were mean centered by subtracting the mean of each variable. The mean centered independent variables were then used in the analysis. Therefore, the VIF scores were lower than 10. The interaction term was calculated by multiplying the mean centered disclosure and mean centered source credibility variables.
Table 4.10 shows the results of the analysis. The model was significant (R2= 0,167, F(5,191)=7,674, p=0,000). H6 is rejected because the interaction term failed to show significance (B=,064, t=,319, p=0,750). There was a marginally significant, positive, direct effect of source credibility on purchase intention (B=,240, t=1,959, p=0,052) as already established in model 3. Disclosure did not show a significant effect (B=,143, t=,814, p=0,417) but the control variable product involvement had a significant positive effect (B=,260, t=4,188, p=0,000) and the second control variable attitude towards the influencer had a marginally significant positive effect (B=,152, t=1,733, p=0,085) on purchase intention. An additional linear regression without the control variables was performed (R2= 0,082, F(3,193)=5,764, p= 0,001). Nevertheless, there is no support for H6 since the interaction term was not found to be significant (B=,065, t=,311, p=0,756).
4.2.7 Testing the Conceptual Model
In a last step, the previous separately tested hypotheses H1 to H6 were assessed together using Hayes PROCESS model 58 with a number of 5000 bootstrap samples and 95% confidence intervals. The independent variables were mean centered since model 6 showed VIF scores over
Table 4.10: Results with purchase intention as dependent variable (model 6; standardized coefficients)
Model 6
Main Variables
Source Credibility ,162c
Control Variables
Product Involvement ,281a
Attitude towards the Influencer ,139c
Moderator Disclosure ,054 Interaction Effect Source Credibility*Disclosure ,022 R2 (Adjusted R2) ,167 (,145) R2 change ,167 F-value 7,674a
Note: a p-value < ,01; b p-value < ,05;
32
10 (Hair et al., 2014). The mean centered independent variables were then used in the analysis and thus the VIF scores were below 10. Product involvement and attitude towards the influencer were used as control variables since they were found to be significant on the concluding dependent variable purchase intention. The outcome of the analysis can be found in Figure 4.1.
Note: a p-value < ,01; b p-value < ,05; c p-value < ,10
Figure 4.1: Results with Hayes PROCESS model 58
33
significant, direct, positive effect on source credibility (B=0,404, t=8,990, p=0,000), establishing that a more positive attitude towards the influencer leads to a higher perceived credibility of the influencer. The only difference to the separate analysis is that the control variable attitude towards the influencer was not found to have a positive, marginally significant effect on purchase intention (B=0,125, t=1,522, p=0,130) as it was established in model 3 and 6.
The same analysis was conducted without the covariates. Again, H1, H2, H4, H5 and H6 get rejected and H3 is supported which confirms the previous analysis. In addition, two interesting findings were noted. First, for H2 the type of influencer now had a marginally significant effect on source credibility (B=-0,222, t=-1,770, p=0,078), but with a negative sign, in contrast to the prediction. This indicates that the traditional influencer is perceived as more credible than the non-traditional influencer which leads to the rejection of H2. Second, the positive effect of source credibility on purchase intention was now highly significant (B=0,392, t=3,686, p=0,000) compared to the marginally significant effect when the control variables were included. This indicates that a lot of variance is explained by the control variables. In summary, these findings are in line with the previous analysis.
4.3 Additional Analyses
Three additional analyses were conducted. Therefore, the single dimensions of source credibility namely expertise, trustworthiness and attractiveness were used separately as mediator in the Hayes PROCESS model 58 with a number of 5000 bootstrap samples and 95% confidence intervals (see table 3.2 for factor loadings and Cronbach’s alpha). The type of influencer was used as independent variable, disclosure as the moderator, purchase intention as the dependent variable and product involvement as well as attitude towards the influencer as covariates. Because of VIF scores over 10, the independent variables were mean centered which led to VIF scores below 10 (Hair et al., 2014). The mean centered variables were then used in the analyses. The goal of the additional analyses was to investigate the role of source credibility in more detail since H3, the positive effect of source credibility on purchase intention, is the only confirmed hypothesis.
34
on purchase intention (B=0,308, t=3,644, p=0,000) which shows that the more trustworthy the influencer appears, the higher the purchase intention of the Instagram user will be. In addition, a positive, significant, direct effect on trustworthiness of disclosure compared to non-disclosure was found (B=0,390, t=2,655, p=0,009). This indicates that using disclosure in a post increases trust towards the influencer. The dimension “attractiveness” did not show a significant effect on purchase intention (B=-0,141, t=-1,609 p=0,109) which shows that Instagram users do not take the influencer’s attractiveness into account for their purchase intention. In addition, it was found that the traditional influencer is perceived as more attractive than the non-traditional influencer (B=-0,329, t=-2,255, p=0,025). Also, there was a significant interaction effect of disclosure and the type of influencer (B=0,622, t=2,187, p=0,030) which was only significant for the traditional influencer (p=0,002) and not for the non-traditional influencer (p=0,955). This indicates that the traditional influencer is seen as more attractive when no disclosure is used compared to a post with disclosure. The interaction graph of the type of influencer and disclosure can be found below in figure 4.2.
35
5 Conclusion
5.1 Discussion
The goal of the thesis was to generate insights into the moderating role of disclosure in influencer marketing on Instagram when different types of influencers are used. More specifically, it was investigated if disclosure strengthens the relationship between the type of influencer and source credibility and if disclosure on the other hand mitigates the effect of the influencers’ credibility on purchase intention. In addition, it was analysed how the type of influencer effects purchase intention, either directly or through the influencers’ credibility. The results of each hypothesis analysis, as found when Hayes PROCESS model 58 was used, can be found in table 5.1. The discussion of the hypotheses results, the effects of the control variables and the additional analyses is split into two parts. First, the design of the study is reflected on and second, substantive explanations are given.
Table 5.1: Results of hypotheses
Hypotheses Results
H1: Non-traditional influencers have a larger positive effect on purchase intention than traditional influencers.
Rejected
H2: Non-traditional influencers are perceived as more credible than traditional influencers.
Rejected
H3: Source credibility has a positive effect on purchase intention. Confirmed H4: The effect of the influencer type on purchase intention will be
partially mediated by source credibility.
Rejected
H5: Disclosure enhances the effect of type of influencer on source credibility.
Rejected
H6: Disclosure mitigates the effect of source credibility on purchase intention.
Rejected
5.1.1 Design of the Study
36
disclosure lasts, the stronger its effect was found to be (Boerman et al., 2012). In some cases, it might even be that respondents did not notice the disclosure.
Second, using the “@Frish” tag in the stimulus posts might have led to a non-significant effect of the influencer type on credibility (H2). An important aspect in order to be perceived as credible is the authenticity of the influencer (Audrezet, de Kerviler, & Guidry Moulard, 2018). Therefore, the reason for not finding a connection between the influencer type and source credibility might be the low authenticity of the stimulus posts used in the survey. Audrezet et al. (2018) made a distinction between two kinds of authenticity. First, passionate authenticity represents the non-commercial interest of the influencer. Second, transparent authenticity describes the disclosure of the partnership. In all four conditions the commercial interest was visible by tagging the brand “@Frish” which could lead to a low passionate authenticity. In addition, the transparent authenticity was decreased in the posts without disclosure.
Third, only one representative of each category of influencer type was presented as a stimulus. Thus, results could be tied to the particular personas Selena Gomez and Zoella. Specifically, the following findings should be considered with caution. A marginally significant effect of the type of influencer on source credibility was found, when the control variables were removed (H2). Notably, the effect is different than predicted. The traditional influencers are seen as more credible than the non-traditional influencers. Also, the additional analysis of the attractiveness dimension led to new insights. It was indicated that traditional influencers are perceived as more attractive than non-traditional influencers and that disclosure weakens this effect for traditional influencers. Future research should use multiple influencers of one category to confirm these findings.
5.1.2 Substantive Explanations
Literature is taken into account to give substantive explanations. In the following, literature is used to confirm significant findings on the one hand and on the other hand to explain insignificant and additional findings.
37
Atkin and Block (1983) might offer an explanation to why there was no direct effect of the type of influencer on purchase intention (H1). They argued that in order to achieve a reaction to a product, the image of the endorser has to be linked with the brand. Therefore, the cause of the insignificant finding could be that the respondents of the survey did not have an existing image of the influencer to connect with the brand.
Partial mediation of source credibility between the type of influencer and purchase intention was not established (H4). The relationship was assumed because consumers associate attributes like credibility with an influencer and those in turn effect their final response to the brand (Amos et al., 2008). Since source credibility is not the variable explaining the underlying mechanism, it could be that not attributes connected to the influencer, but attributes of the followers mediate the relationship. Persuasion knowledge for example is often used in advertising literature when disclosure is involved (Boerman et al., 2012, 2017). Therefore, different types of influencers could activate persuasion knowledge in distinctive ways which finally effects purchase intention.
No moderation effect of disclosure on the relationship between the type of influencer and source credibility was found (H5). One argument was that a positive relationship between the non-traditional influencer and the follower should be established (Carl, 2008). Thus, a negative or missing relationship with the non-traditional influencer could be the cause for insignificant findings.
38
sources cause behavioral intentions like purchase intention (Sternthal et al., 1978). Furthermore, it was established that disclosure can have a positive effect on trustworthiness (Chapple & Cownie).
Last, the additional significant findings concerning the control variables are discussed. The control variable product involvement showed a positive effect on source credibility as well as on purchase intention. This indicates that with an increasing product involvement, the influencer is evaluated as more credible and the purchase intention for “Frish” rises. Literature stated that depending on the level of involvement, different routes of information comprehension are taken by a person. This concept is generally known as the Elaboration Likelihood Model (Petty & Cacioppo, 1986). Specifically, it was found that when the involvement is low, information is not analysed in detail. If the involvement is high, in-depth evaluation is done (Maheswaran & Meyers, 1990). This could explain the positive effect of product involvement on source credibility and purchase intention. When the product involvement of the respondents is low, they only have a quick look at the stimulus post and do not reflect on the given information. This might lead to a low credibility and low purchase intention. If followers have a high product involvement, they analyse the stimulus post in-depth and look for example at the capture, number of likes and the product presentation. This more thorough reflection might convince the subjects that the influencer is a credible source and might increase the chance of finding arguments for a purchase.
The second interesting control variable is the attitude towards the influencer. It was found to have a positive effect on source credibility, indicating that the more positive the attitude towards the influencer is, the more credible the influencer appears. Hoyer and MacInnis (2008) defined attitudes as a lasting assessment of a person that indicates the degree of liking or disliking. The established attitudes then effect our thoughts. Therefore, it is possible that people that have a positive attitude towards an influencer also think of the influencer as more credible and vice versa since the established attitudes influence the subsequent thoughts.
5.2
Management Implications
39
intention of Instagram users, cooperating with influencers who have a high credibility should be considered. This can be done by finding influencers towards which the marketer’s target group has positive attitudes. Also, advertisement should be done for products that potential customers are highly involved with. This can create high credibility for influencers and can increase purchase intention.
When looking at the dimensions of an influencer’s credibility in detail, further implication can be given. Trustworthiness is identified as the dimension that increases purchase intention. Therefore, it is important that influencers that appear trustworthy to Instagram users are considered as brand endorsers. Interestingly, trustworthiness can be increased when disclosure is used in a post. This indicated that transparency about the sponsorship, in this specific situation, is important when conducting an influencer marketing campaign. When marketers want to work with an attractive influencer, they should concentrate on traditional influencers since they are perceived as more attractive than non-traditional influencers. However, they should keep in mind that in this particular setting, disclosure mitigates the effect of the traditional influencer on attractiveness.
5.3 Limitations and Future Research
The research comes with its limitations. Firstly, the sample size of 197 respondents is relatively small which could lead to false inferences because the sample might not accurately represent the whole population (Malhotra, 2010).
Secondly, since the survey was mainly distributed through the personal network, it can be assumed that many participants’ native language is not English but German or Dutch. This could lead to wrong answers caused by difficulties in understanding the questionnaire. Controlling for the level of English of the participants could have given more insights. In addition, recording from which country the participants are could as well have been added as a control variable. Research found that social media’s impact on purchase intention differs in culture (Goodrich & De Mooij, 2013).
40
Therefore, based on the limitations and findings of the thesis, future research directions are suggested. Firstly, to find out if there is generally no effect of disclosure as a moderator on the two predicted relationships as well as no effect of the type of influencer on source credibility and purchase intention, the research should be repeated with multiple different stimuli. Instead of one representative per category of the influencer type, multiple representatives per category should be used. They might differ in gender, age and in the fields through which they became famous like sports or cooking. In addition, multiple product categories like fashion, food and fitness should be included. Also, a distinction between fictitious and well-known brands could be made.
41
6 Literature
Amos, C., Holmes, G., & Strutton, D. (2008). Exploring the relationship between celebrity endorser effects and advertising effectiveness: A quantitative synthesis of effect size.
International Journal of Advertising, 27(2), 209–234.
Association of National Advertisers (2018, April 3). Advertisers Love Influencer Marketing:
ANA Study. Retrieved September 10, 2019, from https://www.ana.net/content/show/id/48437
Atkin, C., & Block, M. (1983). Effectiveness of Celebrity Endorsers. Journal of Advertising
Research, 23(1), 57–61.
Audrezet, A., de Kerviler, G., & Guidry Moulard, J. (2018). Authenticity under threat: When social media influencers need to go beyond self-presentation. Journal of Business Research, 1-13.
Baron, R., & Kenny, D. (1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of
Personality and Social Psychology, 51(6), 1173-82.
Baxter-Wright, D. (2017, September 13). Zoella's career timeline: How the YouTuber became
a household name. Retrieved September 21, 2019, from
https://www.cosmopolitan.com/uk/entertainment/a12227772/zoellas-career-timeline/ Bearden, W. O., Netemeyer, R. G., & Teel, J. E. (1989). Measurement of Consumer Susceptibility to Interpersonal Influence. Journal of Consumer Research, 15(4), 473–481. Bergkvist, L., & Zhou, K. Q. (2016). Celebrity endorsements: A literature review and research agenda. International Journal of Advertising, 35(4), 642–663.
Biography. (2019, October 24). Selena Gomez. Retrieved December 6, 2019, from https://www.biography.com/musician/selena-gomez
Boerman, S. C., Van Reijmersdal, E. A., & Neijens, P. C. (2012). Sponsorship Disclosure: Effects of Duration on Persuasion Knowledge and Brand Responses. Journal of
Communication, 62(6), 1047–1064.
———, ———, and ——— (2013). Appreciation and Effects of Sponsorship Disclosure,
Advances in Advertising Research IV: The Changing Roles of Advertising, Sarah Rosengren,
42
———, ———, and ——— (2014). Effects of Sponsorship Disclosure Timing on the Processing of Sponsored Content: A Study on the Effectiveness of European Disclosure Regulations. Psychology and Marketing, 31(3), 214–24.
———, ———, and ——— (2015). Using Eye Tracking to Understand the Effects of Brand Placement Disclosure Types in Television Programs. Journal of Advertising, 44(3), 196–207.
Boerman, S. C., Willemsen, L. M., & Van Der Aa, E. P. (2017). “This Post Is Sponsored”: Effects of Sponsorship Disclosure on Persuasion Knowledge and Electronic Word of Mouth in the Context of Facebook. Journal of Interactive Marketing, 38 (May 2017), 82–92. Bräutigam, F. (2019, March 21). Auch das Private ist geschäftlich. Retrieved October 15, 2019, from https://www.tagesschau.de/inland/urteil-influencerin-pamela-reif-101.html
Campbell, M.C., & Kirmani, A. (2000). Consumers’ use of persuasion knowledge: The effects of accessibility and cognitive capacity on perceptions of an influence agent. Journal of
Consumer Research, 27(1), 69–83.
Campbell, M. C., Mohr, G. S., & Verlegh, P. W. J. (2013). Can disclosures lead consumers to resist covert persuasion? The important roles of disclosure timing and type of response.
Journal of Consumer Psychology, 23(4), 483–495.
Carl, W.J. (2008). The role of disclosure in organized word-of-mouth marketing programs.
Journal of Marketing Communications, 14 (3), 225-241.
Carr, C.T. & Hayes, R.A. (2014). The Effect of Disclosure of Third-Party Influence on an Opinion Leader's Credibility and Electronic Word of Mouth in Two-Step Flow. Journal of
Interactive Advertising, 14 (1), 38-50.
Chandran, S. & Morwitz,V.,G. (2005). Effects of Participative Pricing on Consumers’ Cognitions and Actions: A Goal Theoretic Perspective. Journal of Consumer Research, 32 (September), 249-259.
Chandrasekaran, R. (2004). The Influence of Redundant Comparison Prices and Other Price Presentation Formats on Consumers’ Evaluations and Purchase Intentions. Journal of Retailing, 80 (1), 53-66.
Chapple, C., & Cownie, F. (2017). An Investigation into Viewers’ Trust in and Response Towards Disclosed Paid-for-Endorsements by YouTube Lifestyle Vloggers. Journal of
