Influencers Era: what makes their endorsements
more persuasive?
Experimental analysis of the effects of influencers’ attractiveness and
trustworthiness on purchase intentions, and underlying processes.
Nicole Menasce Rosset 11840773
Supervisor: Dr. Stephanie Welten
Persuasive Communication
Graduate School of Communication University of Amsterdam
Master’s Thesis 31-01-2020
Abstract
Social media apps such as Instagram have been lately through a massive growth, not just on the number of users, but also when it comes to different strategies of monetization of the platform. Companies saw on the rise of Instagram influencers an opportunity to speak to target groups throughout a source that consumers have an involvement with. These new celebrities have demonstrated impressive results when it comes to product endorsements, but comprehending what makes them more persuasive in a scientific matter has become crucial. Besides the commercial goal, this research performed an experimental design considering ninety women (N = 90) with ages ranging from 18 to 44 years old to understand the causal effects of the influencers’ source characteristics (attractive vs. trustworthy) on purchase
intentions. It was also hypothesized that there are underlying processes occurring on the
platform users that might affect the intentions that consumers have. In this present research,
wishful identification and parasocial interaction were analyzed as mediators of the
relationship between the source characteristics and purchase intentions. Among the participants of this study, it was found that there is no direct effect of the source
characteristics on purchase intentions, neither any of the parallel mediation of wishful identification and parasocial interaction was found to be significant. Furthermore, this
research instigated reflections to be addressed on future research mainly concerning the possibility of a manipulation of source characteristics.
Introduction
Online influencers are a new channel of advertising that was originated by social media apps such as Instagram. These new celebrities have often been used by brands on their marketing strategies, mainly because of their audience reach and engagement (easier to work with niches and specific targets), original branded content, and more importantly for this research, the consumers trust on them (Aguiar & Reijmersdal, 2018).
The understanding of how and what makes these influencers more persuasive has been explored lately in the communication research field. This is not just because of the practical relevance in a commercial matter of how to make these endorsements more successful, but also because it helps to explain new types of relationships created between people in the online world.
A new space for product endorsement was created, and brands look for these influencers to be their brand ambassadors to create long-term partnerships (Aguiar &
Reijmersdal, 2018). Traditional celebrities have been used for a while as marketing strategies, mainly because of testimonial marketing, when a consumer talks about the product as a first-person's perspective like a review (Ohanian, 1990). Ohanian (1990) explained that this strategy is effective because of the perceived similarity people feel with the endorser. However, these traditional celebrities (e.g. actresses, singers) lack on something that Instagram influencers have: they tend to address their followers on a personal level on their posts, giving the impression of closeness, allowing the followers to see them as peers
(Schouten, Janssen & Verspaget, 2019). Also, the possibility to interact and comment on the influencers post strengthens the idea of similarity and proximity (Schouten et al., 2019).
Even though research concerning the effectiveness of online influencers on purchase
intentions of endorsed products has been conducted, there is no current research in how the source characteristics (attractiveness vs. trustworthiness) of these new celebrities are
manipulated and analyzed in combination with the underlying processes that might happen on the viewer's mind when exposed to the endorsement (wishful identification and parasocial
interaction). Previous studies mainly focused on the direct effects of source characteristics
on purchase intentions or how differently traditional celebrities are perceived from Instagram influencers on their attractive or trustworthy characteristics. Additionally, these studies were predominantly composed of surveys, which resulted in the lack knowledge of the causal effects of the variables on the outcomes. This research is primarily focused on performing an experiment in which the manipulation of the independent variable source characteristic (attractiveness vs. trustworthiness) is tested and if underlying processes such as the mediation of the variables wishful identification and parasocial interaction affect the dependent variable purchase intentions. The establishment of a causal effect of these variables on purchase intentions can rule out other possible influences that the previously conducted surveys might have considered.
Wishful identification and parasocial interaction can be underlying processes that can
occur when the viewer is presented to a trustworthy or attractive influencer. Parasocial
interaction can be associated with a more trustful influencer because, to be able to create this
"friendship like" relationship with him/her, this viewer needs to trust him/her. On the other hand, wishful identification can be associated with the attractive influencer, because it is related to the "wish to be like" the person being portrayed.
Comprehending how to make influencer marketing the most profitable on sales prossible (by using purchase intentions as the outcome variable) has a big practical relevance, primarily because this strategy has gained a lot of attention in recent times, and also because it generates big investments in this field of marketing. Consequently, the research question of this present study is:
RQ: “How does trustworthiness compared to attractiveness of an Instagram Influencer affect the purchase intention on endorsed products, and to what extent do Parasocial Interaction or
Wishful Identification that the consumer has with the Influencer explain these effects?”
Theoretical Framework
With the development of new medias such as social medias, all kinds of new forms of relationships were created, the same way brands had to also adapt their way to communicate with their consumers. Instagram became an extremely useful tool for these companies to reach targeted audiences that are not easily reachable by traditional media or advertising (Aguiar & Reijmersdal, 2018), especially when using Instagram influencers to endorse the products. These new celebrities are interesting in a commercial perspective because they are seen as more credible and relatable, which makes them even more powerful than traditional celebrities when it comes to persuasion (Djafarova & Rushworth, 2017). Aguiar and
Reijmersdal (2018) also explain that these social media influencers usually have an
advantage over other celebrities as they have created their own personal brands, and usually this is followed by some sort of specialty.
The differentiation between traditional celebrities and Instagram influencers has been studied and strategies concerning when to use each one of these endorsers were traced according to the different goals companies might have (Schouten et al., 2019). However, the understanding of the influencer's field as its self has yet to be explored to its full potential. explored that much yet. In other words, Instagram influencers have been looked at as a homogeneous source, and this study intends to dissect in more depth the differences between the distinctive types of influencers and their characteristics.
Source credibility is a concept often used to identify among marketeers which
effect on the persuasiveness of the message (Ohanian, 1990). This concept can be applied to any considered celebrity in this context and is defined by three dimensions: trustworthiness,
expertise, and attractiveness (Lafferty & Goldsmith, 1999). All three dimensions construct
the general concept, but they can be studied individually since they are distinctively different.
Source credibility is known to influence persuasion (Lafferty & Goldsmith, 1999) producing
positive attitudes towards what the influencer, in this case, is endorsing.
The main focus of this study relies on two dimensions of these three that constitutes
source credibility: trustworthiness and attractiveness. These two constructs were chosen as
they explain to its full extent the persuasiveness effect of a message when considering the credibility of it (Ohanian, 1990). As credibility of a source was found to influence persuasion (Lafferty & Goldsmith, 1999), this study aims to understand which one of those two
constructs influences more the persuasiveness of an endorsed product (on purchase
intentions), when posted by an Instagram influencer.
Source characteristics is the independent variable of this study and this factor is
constructed by two items: attractiveness and trustworthiness. Trustworthiness of a source concerns how much the public trusts the one who is talking to them and if they perceive this person as someone with enough knowledge on the subject (Ohanian, 1990). It is the degree of confidence and the level of acceptance that the listener has on the person transmitting a message.
According to Ohanian (1990), celebrities who are liked are also most likely trusted. This findings proved that the more trustworthy the communicator, the more persuasive he/she will be, and this trust can be measured by items such as the perceived honesty, reliability, and other constructs by the viewers towards the influencer (Ohanian, 1990). It is important to understand that to be trusted, the endorser must transmit an image in which they posses knowledge of what they are communicating to their followers, being perceived as reliable for
the specific topic in discussion, allowing the customer to rely on that person when it comes to said subject. But also, this influencer has to be liked by the viewer, because if there is no type of likeability, the trust won't be that strong (Ohanian, 1990).
Attractiveness - the other item that builds the factor source characteristics - covers how much the audience perceives that endorser as someone attractive. Facial and physical
attractiveness are what constitutes this item. According to Ohanian’s (1990) study, this is an
important cue when it comes to the initial judgment of someone. But, it is not just about physical appearance, it is also about "chicness" and other physical aspects that are not just part of the physiognomy (Ohanian, 1990). It is interesting to notice in this concept the fact that this endorser is more liked is because of hers/his image. It can be the perception of their personality or/and their physical appearance, but it has more to do with an imagery created than necessarily the professionalism or the authority about the subject (Ohanian, 1990). Attractive (vs. unattractive) communicators were found to be more liked and have a positive impact on products they were associated with (Ohanian, 1990).
In the study by Schouten et al. (2019), the authors explain that influencers can be perceived as more relatable than the traditional celebrities because of the possibility of the influencer addressing the followers directly on their posts, or even the fact that the followers can comment and perhaps interact with this influencer. The fact that there is a "real"
interaction, that the follower is indeed responding or being answered by someone they admire, make these new celebrities more approachable, thus more trustworthy (Schouten et al., 2019). Besides that, influencers are already known for sharing user-oriented product reviews, as well recommend and share personal experiences, and in the consumers’ mind, this content created by influencers reflects honest opinions (Schouten et al., 2019). Even though Instagram users cannot be sure of the authenticity of what is posted by these
influencers, the idea that they are sharing their day-by-day makes them more relatable than traditional celebrities, thus, more trustworthy (Aguiar & Reijmersdal, 2018).
The main question that remains when it comes to analyzing previous studies is to find out which one of these source characteristics (attractiveness vs. trustworthiness) leads to higher purchase intentions. Even though it was found that attractiveness leads to a bigger likability of the source (Ohanian, 1990), trustworthiness might have a greater influence on persuasion because, according to Aguiar and Reijmersdal (2018), trust is one of the main four reasons to use influencer marketing with a persuasive goal. The other three reasons are "engagement", "reach", and "original content". It is possible to understand from these four reasons of why to use influencer marketing, that if trust is part of the items, and the
attractiveness of the influencer is not, being trusted by the followers might have greater
impact on marketing goals such as purchase intentions. Ohanian (1990) also states that celebrities who are liked will also be trusted. Aguiar and Reijmersdal (2018) also comment the fact that relatability with the influencer gives the follower this feeling of "proximity", which leads to a bigger trust on them. Since "relatable" is one of the items that constitute the scale to measure trustworthiness, the first hypothesis is:
H1: A trustworthy influencer leads to stronger purchase intentions of an endorsed product
compared to an attractive influencer.
Involvement with characters is a concept that explains different relationships created between spectators and someone communicating to an audience (being directly or indirectly) (Giles, 2002). Two specific types of involvement will be discussed in this study:
parasocial interaction and wishful identification. These two types were chosen because of its
relationship with the independent variables of the study (trustworthiness and attractiveness). The main assumption in this study is due to the fact that parasocial interaction might mediate
the relationship between the purchase intention and the trustworthiness, and wishful
identification might mediate the relationship between purchase intention and attractiveness. Parasocial interaction can be described as "companionship" and "personal
identity" created between the spectator and the character (Giles, 2002). Such as in soap opera novels, viewers can relate to the character as someone that could be in their social circle, that might remind them of people that they already know, creating a pseudo-relationship (Moyer-Gusé, 2008). This is an important aspect of the Parasocial interaction because the viewer sympathizes with the character as someone they could be friends with, as a companionship relation (Ohanian, 1990). This might be related to the trust that this character (in the case of this study the influencer) passes to their viewers. Chung and Cho (2017) found in their study that parasocial interaction mediated the relationship between social media interactions and source trustworthiness, and that this trust had a positive effect on brand credibility, leading them to understand that influences as well people's purchase intentions. Hoffner (1996) also discovered that parasocial interaction increases viewers willingness to accept information from media personalities, in this case, Instagram influencers.
Wishful identification, on the other hand, can be defined as the desire of the
viewer to be like or behave in a similar way to the character presented, wishing to be more alike (Giles, 2002). Usually we wish to be like people we admire somehow, thus if the influencer is perceived as attractive, this might lead to a Wishful Identification with her/him. This desire to “be like” is different from the parasocial relation in the sense that brings a lot more of idealization of the character, and not the almost close relationship that the latest mentioned brings to the spectator. Liking and feeling "close" to the endorser as a "friend" doesn't mean that the viewer wishes to be like them. Not necessarily if you are friends with someone you desire to be like them, but if you admire someone's style, clothes, or posture and want to replicate this on you, then you need to be attracted to that imagery created by that
influencer somehow. Hoffner (1996) explain in her study that for women characters, physical attractiveness predicted the desire to be like (i.e. wishful identification) for girls. Which can be then related to the attractiveness perceived by the viewers of an influencer.
An interesting side of influencer marketing that can be approached here is the fact that it might be possible to define the influencers' content as a construction of a storyline. What they post every day, what they expose, say about it can all be considered a construction of a narrative. As explained by Moyer-Gusé, Jain, and Chung (2012), a narrative structure has an identifiable beginning (the start of the influencer's career), middle and end (which can be considered all that has been done until now), and provides information about the scene, characters, and conflict (being the content of the posts in general). This aspect of influencers is interesting for this study because it relates to the high persuasiveness of advertising that uses this tool of storytelling. The usage of narratives structure can reduce resistance to persuasion (Moyer-Gusé et al., 2012), making the followers of that influencer more open to receiving the endorsement and be persuaded by it.
Moyer-Gusé et al. (2012) explains that the engagement with the dramatic factor of a narrative of the viewers leaves them in a less critical state, less likely to counter argue with the embedded persuasive message, which makes it easier to influence behaviors. Thus, the fact that Instagram influencers constitute a daily "story-telling" of their lives is also an argument of why they can be considered more persuasive. In regard to this persuasive effect, Moyer-Gusé et al. (2012) discusses that involvement with the characters in the narrative is one of the main components of this engagement.
Parasocial interaction can also relate their persuasiveness to the fact that
influencers create a narrative structure on their social media. The main association here between these two concepts is due to the fact that this almost personal relationship between the character (influencer) and the audience, there is less resistance to persuasion and a
character (such as an influencer) can decrease the perception of persuasive intention in the first place (Moyer-Gusé et al., 2012). Followers of an influencer might be from the start more open to listening their tips because of this relationship created between them.
Because of the arguments above, the second hypothesis is:
H2a: The effect that a trustworthy influencer leads to stronger purchase intentions of an
endorsed product compared to an attractive influencer, will be positively mediated by parasocial interaction.
Even though this present study believes that the trustworthy condition of this study will lead to stronger purchase intentions when compared to an attractive influencer, when positively mediated by a parasocial interaction relationship created between the viewer and the influencer, an attractive influencer might as well have an effect (just not as strong as the
trustworthiness one). Throughout the analyses for references to this present study, repeatedly
it was found that both trustworthy and attractive characters were influential and persuasive, however the connection between the concepts used generated the hypothesis that
trustworthiness is more effective than attractiveness. However, this first hypothesis does not
exclude the effect that attractiveness might also have on purchase intentions. Thus, the second part of the second hypothesis is:
H2b: The effect that a trustworthy influencer leads to stronger purchase intentions of
an endorsed product compared to an attractive influencer, will be negatively mediated by wishful identification.
This study aims to understand how different elements of source characteristics (attractiveness vs. trustworthiness) influence the purchase intentions of an endorsed product by an influencer and if the parasocial interaction or wishful identification relationship that the consumer has towards the influencer can explain the relationship.
Also, this specific study has a scientific relevance because it can examine involvement such as wishful identification as a direct mediator for attractiveness, and
parasocial interaction as a mediator of trustworthiness, and how these both mediated
relationships lead to higher purchase intentions. These underlying processes haven’t been studied and compared this way. Also, the fact that this study is operationalizing the scales from Ohanian (1990) is also an innovative aspect of this research, meaning that until now there was no causal effect testing of the source characteristics on purchase intentions.
There is a certain urgency into understanding human behavior on online platforms because of its increasingly usage and direct influence into people's lives and day by day. The sooner we understand how people respond to different stimulus online, the better we can understand the outcomes on the "real world".
Figure 1. Conceptual Model
Methods
Participants and Design. A total amount of one hundred and forty-six people participated in this experiment. These participants were randomly assigned to either one of the two conditions of a one factor design: source characteristics (attractive vs. trustworthy) consisting on an online experiment. From this amount, a total of fifty-five cases were deleted because of different reasons, resulting in a total of ninety participants (N = 90). Ten
respondents were deleted because they were male, and for this research only female participants were required. One participant was excluded because she did not meet the
minimum level of English proficiency (B1), marking down an Elementary Level (A1). This is important because the participants need to understand English above the level of B1
proficiency to make sure that this the person understands not only the manipulation videos, but also the questions being used in the study in the survey. People with understandings of English with a level of B1 or above proficiency are able to comprehend the information needed from them in this study (Tracktest.eu). From all Instagram users, women between 18 and 44 years old represent 40,1% of these users according to the website Statista (Clement, 2019), and one participant was excluded because she is fifty-seven years old, which is above the age range that was required. The other forty-three excluded respondents didn’t finish completing the survey; thus, they were not considered. The final sample consisting on ninety participants consists on females between 19 and 44 years old (M= 26.4, SD= 4.71) and their English Proficiency level varied between B1 (Intermediate English) and C2 (Proficiency English) the scale going from 5 (B1) till 8 (C2) (M= 7.27, SD= 0.76).
The participants sample were recruited based on the author's personal reach, sent to people on social media, message and WhatsApp broadcasting and email. The title of the study is "Online Influencers Differences", and the message sent to recruit participants addressed the fact that for this study, women between 18 and 44 years old were needed, that
the filling of the survey would take about 10 minutes, and, as an emotional appeal, this answers would help the author graduate her master.
Procedure and Manipulations. The experiment consisted of two videos in which either attractiveness or trustworthiness was manipulated based on the condition of the independent variable source characteristics. Both videos were a fictional representation video of an Instagram story; that is a tool commonly used by influencers to endorse products (Aguiar & Reijmersdal, 2018). One of the videos had the goal to represent a trustworthy influencer, the other an attractive.
Since no previous study that manipulated trustworthiness was found, the items from the Ohanian (1990) scale were used as a guideline to create the lines and body language of the actress in the trustworthiness video condition. Some of the items were “Honest” and “Sincere” (see Appendix for the whole scale). From the 5 items that constitutes this scale, the script of the video was created making sure that these characteristics could be noticed by the participants. The idea of the script was based on the attempt to create a way to manipulate
source characteristics, making sure that the perception of the viewer changed depending on
how the character was perceived. The lines and the gestures of the actress were created based on the items from the Ohanian (1990) scale for trustworthiness measure, using some of the words on it during her speech, a body language that transmitted an idea that she was
accessible and reliable. The usage of lines such as “trust me” had the intention of giving this
trustworthy image.
The attractiveness video condition had the intention of portraying an attractive influencer, and the same scale from Ohanian (1990) was used to create this script as well. Some of the 5 items that constitute the scale are “Sexy” and “Classy” (see Appendix for the whole scale). The same way the trustworthiness video script was built, the attractiveness one was; the physical appearance of the influencer was modified from the first condition
mentioned to more makeup and accessories, targeting the items such as "Classy" and "Elegant", since she was wearing a blazer in this condition and in the trustworthy one a t-shirt. Part of the goal of this characterization was making sure that the two video conditions were almost opposite from each other in aspects that could relate to the scale from Ohanian (1990). Since in the trustworthy condition the body language was focusing on appearing accessible, in the attractive condition the script made sure to focus on a body language that transmitted an elegant and sexy image. Unfortunately, for none of the two video conditions of
source characteristics were found studies that manipulated this variable, thus this study had
the intention of doing so based on the scale from Ohanian (1990).
Each participant watched one of the videos and then answered questions concerning the variables in the study, some demographic questions, and two control questions that had the goal of making sure the randomization was well done.
Manipulations checks for Source characteristics. The first set of questions were concerning the independent variable source characteristics.
Attractiveness was measured with five items from Ohanian (1990) on a 7-point Likert
scale, in which the respondents had to rate how they perceived the influencer in the manipulation video within two opposite words such as “Not Classy” and “Classy”, and “Ugly” and “Beautiful” (see Appendix for all the 5 items). The lowest degree started at 1 (e.g. “Ugly”) and the highest value was 7 (e.g. “Beautiful”). According to the factor analysis, these 5 items explain 68.3% of the variance in the perception of the influencer’s
attractiveness. The items form with factor loadings above .79 and an Eigenvalue of 3.41 one
component. The reliability test showed that these items form a reliable scale (α = .88) and no item must be deleted.
Trustworthiness was measured as well with five items from Ohanian (1990) on a
the manipulation video with two opposite words such as “Dishonest” and “Honest”, and “Insincere” and “Sincere”(see Appendix for all the 5 items). The lowest degree started at 1 (e.g. “Dishonest”) and the highest value was 7 (e.g. “Honest”). The factor analysis of this variable containing 5 items results in 71.20% of the variance in the perception of the influencer’s trustworthiness being explained by these items. The items form with loadings above .19 and an Eigenvalue of 3.56 one component. The reliability test showed that these items form a reliable scale (α = .88) and there is no need to delete any item.
After doing the factor and reliability analysis, a One-Way ANOVA was conducted to find out if the attractiveness and trustworthiness perceived by the participants was different among the two video conditions of source characteristics. Concerning the perceived
attractiveness of the influencer, there was no significant difference between the
attractiveness condition (M = 4.47, SD = 1.10) and the trustworthiness condition (M = 3.38, SD = 1.26), F (1, 88) = 0.04, p = .842, meaning that the participants did not perceive the
influencer differently concerning how attractive they thought the influencer was on both video conditions of source characteristics. When it comes to the perceived trustworthiness, there was also no significant difference between the attractiveness condition (M = 4.42, SD = 1.25) and the trustworthiness condition (M = 3.56, SD = 1.26), F (1, 88) = 0.47, p = .497, meaning as well that the influencer was not perceived differently about her trustworthiness on both video conditions.
Mediators. The mediators of the experiment were parasocial interaction and wishful
identification. Parasocial interaction was measured with the scale from Rubin and Perse
(1987) in a 7-point Likert scale in which the participant had to indicate the level of agreement with three different statements, e.g. "Camila Rodrigues makes me feel comfortable, as if I am with a friend." (see Appendix for all 3 items). The scale ranged from “Strongly Agree” (7) till “Strongly Disagree” (1). According to the factor analysis, these 3 items explain 74.36% of
the variance in the parasocial interaction created between the participants and the influencer in the manipulation video. The items form with factor loadings above .68 and an Eigenvalue of 2.23 one component. The reliability test showed that these items form a reliable scale (α = .83) and no item must be deleted. Wishful identification was measured by the scale from Hoffner (1996) also in a 7-point Likert scale regarding the agreement with statements. In the case of this mediator, an example of statement used was "Camila Rodrigues is the sort of person I want to be like myself."(see Appendix for all 3 items). According to the factor analysis, these items explain 76.2% of the variance in the parasocial interaction created between the participants and the influencer in the manipulation video. The items form with factor loadings above .69 and an Eigenvalue of 2.29 one component. The reliability test showed that these items form a reliable scale (α = .83) and no item must be deleted.
Purchase intentions. The dependent variable purchase intention was measured by the scale from Dodds, Monroe, and Grewal (1991) with a 7-point Likert scale. The scale was measuring agreement with statements such as "I am likely to buy this product" (see Appendix for all 3 items), and the scale ranged from “Strongly Agree” (7) to “Strongly Disagree” (1). According to the factor analysis, these 3 items explain 81.33% of the variance in the purchase intention of the participants. The items form with factor loadings above .71 and an
Eigenvalue of 2.44 one component. The reliability test showed that these items form a reliable scale (α = .89) and no item must be deleted.
Results
Randomization check. The randomization check considers the variables: frequency of purchase of beauty products, the Instagram usage, age, and English proficiency level. To test the randomization, an Independent Sample t-test was used for the analysis of the variable concerning the purchase of beauty products habits, and found that the attractiveness
condition (M = 3.55, SD = 0.94) did not significantly differ from the trustworthiness condition (M = 3.67, SD = 0.88) when it comes to the participants beauty product purchase habits, t (88) = -.062, p = .538, 95%CI [ -0.50, 0.26]. When it comes to Instagram usage, a Chi-Square analysis showed a non-significant difference between the usage among the two video conditions, Chi-Square (1) = 0.39, p = .539. Concerning the age, an Independent sample t-test was performed, and the results found were that the attractiveness condition (M = 26.48, SD = 4.98) and the trustworthiness condition (M = 26.33, SD = 4.51) did not significantly differ from each other, t (88) = 0.14, p = .887, 95%CI [ -1.85, 2.13]. Lastly, the English proficiency level at the attractive condition (M = 7.12, SD = 0.86) compared to the
trustworthy condition (M = 7.40, SD = 0.64) had also non-significant differences between
them, t (88) = -1.74, p = .085, 95%CI [ -0.59, 0.39]. Thus, with all these results it is possible to affirm that the randomization of the groups in the conditions attractiveness and
trustworthiness was well done, i.e. there are no differences between them that might explain
the results.
Control variables. Two variables were used as control to make sure that the
randomization of the participants was done in a way that couldn’t distort the results. Thus, the goal of this control variables is to be not significantly different to be able to exclude the possibility of these factors influencing the results. The first variable measured was a question at the end of the survey asking about the frequency in which the person purchased beauty products, and the answers could range from “Always” (5) to “Never” (1). An Independent Sample t-test was used for the analysis and found that the attractiveness condition (M = 3.55,
SD = 0.94) did not significantly differ from the trustworthiness condition (M = 3.67, SD =
0.88) when it comes to the participants beauty product purchase habits, t (88) = -.062, p = .538, 95%CI [ -0.50, 0.26]. Thus, it is possible to consider that this purchase habit is equally distributed among the two groups, excluding the possibility of a distortion of the results since
both groups have similar habits. The other control variable used was about the social media usage of the participants. A list of the most used social medias was presented, and the
participants had to check which they were users of. Since for this study only Instagram usage matters, the analysis will only be done considering the responses for this item. People that checked Instagram as a social media they used were addressed with the value of 2, and people who didn’t check had a zero. Other social medias such as Facebook or Snapchat were assessed in case Instagram usage is not enough to ensure the equal distribution of the sample. With a Chi-Square test it was found no significant difference between the Instagram usage among the two conditions, Chi-Square (1) = 0.39, p = .539. Again, with this result it is possible to say that the participants are equally distributed among the two conditions when it comes to Instagram usage, excluding this as a factor that might impact the results.
Manipulation checks. To understand how the influencer in each condition of the experiment was perceived concerning the independent variables trustworthiness and
attractiveness, and if they differed from each other in each condition, a one-way analysis of
variance (ANOVA) was conducted. The analysis of variance showed no difference between the two video conditions of source characteristics on attractiveness, F(1, 88) = 0.40, p = .842, ƞ2= 0.021, M attractiveness condition= 4.47 (SD = 1.10) , M trustworthiness condition = 3.38 (SD=
1.26). When it comes to the trustworthiness on the two videos of source characteristics, it was also found no difference between the the conditions, F(1, 88) = 0.47, p = .497, ƞ2= 0.07,
M attractiveness condition= 4.42 (SD = 1.25), M trustworthiness condition = 3.56 (SD= 1.26)
Hypotheses testing
H1: “A trustworthy influencer leads to a stronger purchase intention of an endorsed product
compared to an attractive influencer.”.
An Independent Sample t-test was used to compare the purchase intentions of the participants on the two conditions of source characteristics on the study (trustworthy and
attractive videos). The attractiveness condition (M = 3.05, SD = 1.57) did not significantly
differ from the trustworthy condition (M = 3.32, SD = 1.46), t (88) = -0.87, p = .384, 95%CI [ -0.91, 0.35]. Thus, the hypothesis 1 is rejected, since there is no significant difference
between the two source characteristics when it comes to the participants’ purchase
intentions.
H2a: "The effect that a trustworthy influencer leads to stronger purchase intentions of an
endorsed product compared to an attractive influencer, will be positively mediated by parasocial interaction.".
H2b: "The effect that a trustworthy influencer leads to stronger purchase intentions of an
endorsed product compared to an attractive influencer, will be negatively mediated by wishful identification.".
A regression analysis was performed using PROCESS by Hayes (Model 4) to compare if a perceived as trustworthy influencer when mediated by parasocial interaction relationship leads to higher purchase intentions than an attractive influencer mediated by
wishful identification relationship. The first part of the analysis will be done looking into the
relationships between the conditions of source characteristics and the mediators (thus, the mediators are the outcome in this case). The trustworthiness condition of source
characteristics has a positive marginally significant effect on parasocial interaction, B =
0.28, t(0.16) = 1.79, p = .076, 95%CI [-0.03, 0.59]. This means that when the participant is exposed to the trustworthy condition rather than the attractive condition, the parasocial
interaction is positively affected by it. Concerning the attractiveness condition compared to
the trustworthy, a non-significant effect on wishful identification was found, B = 0.35, t(0.26) = 1.35, p = .181, 95%CI [-0.16, 0.86]. This means that the attractiveness video condition had no significant effect on the wishful identification relationship that might have been created by the participant with the influencer compared to the trustworthiness condition.
For the second part of the analysis, three different investigations will be done. Starting with the direct effect of source characteristics on purchase intentions. It was found a non-significant effect, B = 0.14, t(0.31) = 0.47, p = .638, 95%CI [-0.46, 0.75]. Secondly, the direct effect of parasocial interactions on the outcome purchase intentions, was also found a non-significant effect, B = -0.12, t(0.21) = -0.55, p = .583, 95%CI [-0.54, 0.31]. Thirdly, the direct effect of wishful identification on purchase intentions, a positive significant effect was found,
B = 0.48, t(0.13) = 3.70, p < .001 , 95%CI [0.22, 0.74], meaning that wishful identification
positively affects the purchase intention of the participants.
The third and last part of the hypothesis 2 a and b investigation is the indirect effect of the two video conditions (independent variable consisting on trustworthy and attractive conditions) on the purchase intentions (dependent variable). This means that this effect considers the mediators parasocial interactions and wishful identification. The indirect effect considering the wishful identification mediator was found to be non-significant, B = 0.17, 95%CI [-0.08, 0.50]. The effect considering the parasocial interaction mediator was also found to be non-significant, B = -0.03, 95%CI [-0.17, 0.12]. Thus, it is possible to say that the second hypotheses were also rejected.
Figure 2. Conceptual Model with results.
Conclusion and Discussion
Based on theory, it was predicted that trustworthy endorsers would lead to higher
purchase intentions of endorsed products than attractive endorsers (Ohanian, 1990). This
supposition concerns the direct effect of the independent variable source characteristics on the outcome purchase intentions. Results indicated that there was no difference between the two source characteristics conditions when it comes to their effect on purchase intentions, meaning that this hypothesis was rejected. The participants that were exposed to the
attractive condition did not differentiate significantly from the ones exposed to the trustworthy condition about their purchase intentions.
The second hypothesis was divided in two because it concerns the effects of source
with two mediators (wishful identification and parasocial interaction). The hypotheses stated that when exposed to the attractive or the trustworthy conditions, the participants would encounter underlying processes such as wishful identification and parasocial interaction, and these processes supposedly mediate the relationship between the source characteristics and the purchase intentions. According to the literature used to base these hypotheses, wishful
identification would negatively mediate this relationship while parasocial interaction would
positively mediate the effects of source characteristics on purchase intentions. Even though the regression analysis demonstrated that the source characteristics relationship with
purchase intentions were not mediated by any of the predicted mediators (wishful identification and parasocial interaction) - refuting both hypothesis 2a and 2b – other
interesting results about partial effects were found.
The first partial effect found during the analyses of the variables was of that the
trustworthy condition, when compared to the attractive, lead to more parasocial interaction
relationships. The study from Chung and Cho (2017) found that parasocial interaction was a mediator for a trustworthy source when the outcome was concerning brand credibility. But, in the case of this present study, the mediation effect of parasocial interaction was not found, only the partial effect that shows that the trustworthy influencer when compared to the
attractive, a stronger parasocial interaction relationship was created between the participant
and the influencer.
Another interesting effect found (and unexpected when analyzing the theories used to base this study on), is the positively significant effect that wishful identification relationships had on purchase intentions. It was found that the more people created this wish to be like the influencer, the more they intend to purchase the products that she was endorsing. However, a non-significant effect of the parasocial interaction was found on these intentions. According to the study by Moyer-Gusé et at. (2012), the feeling of proximity with a character (in this
case the influencer) leads to lower persuasion resistance, which was interpreted as a bigger possibility of influencing the purchase intentions. Thus, it is possible to see that the results actually gave opposite answers from what was expected at the beginning of this study. In theory, it was expected to find a significant effect of parasocial interaction on purchase
intentions because this relationship variable is based on the feeling of proximity that
Moyer-Gusé et at. (2012) describes. The significant result of wishful identification on purchase
intentions is not necessarily unexpected, but the non-significant effect of parasocial interaction is according to what the theory used in this study argued.
Some elements concerning how this study was conducted, how the manipulation of the experiment was done, and the interpretation of the theories used might explain why these results rejecting the three hypotheses, and the relationships between the mediators and the outcome were found. Also, these results aroused some questions that might require further research to investigate in more depth. The first point to be explored in the discussion of the results found is about how effective the manipulation of the experiment was. It was found that the two video conditions that participants were exposed to were not perceived differently when it comes to their perception of the source characteristics of the influencer. Meaning that if Camila Rodrigues was perceived similarly about her attractiveness and trustworthiness on both video conditions, it is not possible to prove a direct effect that the independent
variable might have had on purchase intentions. The goal to understand which source
characteristics condition would lead to stronger purchase intentions cannot be proven, thus,
it is not possible to affirm that neither the trustworthy condition nor the attractive condition affected differently the purchase intentions of the endorsed product. The main goal of this study was finding a way to manipulate perceived attractiveness and trustworthiness of a character. With these results, it is possible to question if this manipulation is even possible
(since no other study has done so before), or if there are other elements that were not considered in the manipulation of this study that might influence this differentiation.
At the end of the survey, the participants were asked if they had any remarks about what they were impact by. An individual found it hard to be convinced by an unknown influencer (i.e. never saw previous content generated by the influencer presented since she is fictional, and no fake profile was created in this study). This individual believes it is more likely to be influenced by someone that she already follows and knows more about. This is an interesting aspect to be taken into consideration as maybe one of the reasons why the
influencer was not perceived differently on her attractiveness and trustworthiness. Future research should take into account what influencer will be presented to the participants. If a fictional influencer will be portrayed, should a fake profile and a
background of their posts be presented to the participants as well? This way they are able to analyze other aspects of this influencer's profile. Or using a real influencer that is previously known by the participants of the study is a better idea, such as done by Fitri (2018) in her study? This idea can be related to the construction of a storyline discussed by Moyer-Gusé et al. (2012) and the persuasive effect of creating this sequence of events. Thus, possibly the fact that this influencer had no storyline known by the participants, the influencer was not perceived as trustworthy.
The theories analyzed gave strong arguments about the power of a parasocial
interaction relationship on purchase intentions, because of the lower perception of the
persuasive intentions mentioned by Moyer-Gusé et at. (2012) study. Surprisingly though, it was found that the parasocial relationship didn't significantly influence purchase intentions, but wishful identification created by the participants with the influencer did. Could it be possible to state that this could be explained by the lack of previous knowledge on the influencer profile, lifestyle, and personality, making it harder to create a friendship like
relationship (based mostly on trust than physical appearance) than an instantaneous admiration? Similarly, as mentioned above, Moyer-Gusé et at. (2012) study found a limitation concerning the lack of engagement with the characters by the viewers, and that perhaps because of the lack of engagement, a more explicit appeal was required to be persuasive.
Future research should focus on understanding which of these two processes (parasocial interactions and wishful identification) occur first on people's minds, meaning that it should be understood if the timeline of exposure to an influencer affects the
relationship created by the follower with him/her. Does wishful identification occur first as it is considered an initial, superficial judgement? Or does the amount of time in which the follower knows the influencer has an impact on how trustworthy they are perceived?
Wishful identification, differently than parasocial interaction had a significant
positive effect on purchase intentions, and it was the only variable in the study that had this effect on the outcome variable. Can this finding be explained by the fact that a wishful
identification with a character might be faster to occur because it is more related to a
superficial physical perception? However, this desire to be like the influencer was not a result of the source characteristics conditions, meaning that, also differently than predicted,
attractiveness of the influencer did not affect their wish to be like Camila. Thus, another
interesting aspect of this research that should be explored in more depth is understanding what elements increases wishful identification with a character such as an influencer. Understanding how to increase people's wish to be like a character can help brands and the influencer itself to increase the effectiveness of their endorsements and sales. Attractiveness was found to be one predictor of wishful identification on Hoffner (1996) study, but also the intelligence of the character was determinant for the desire to be like among the girls in the
study. For future research, it should be explored in depth what elements contribute better for a strong wishful identification with an influencer.
In order to investigate alternative explanations for the current findings, other
statistical analyses that were not part of the main core of this study were executed. Beginning with the mean value of purchase intentions in general (combining the two video conditions groups). This analysis intended to find out whether there was in fact a meaningful intention to purchase the product being endorsed in general, or if purchase intentions had a ceiling effect occurring, without considering the condition to which the participant were exposed. The result found was that in general, the purchase intentions of all the participants were below the scale midpoint (M = 3.20, SD = 1.51). This aroused the question if the choice of product to be endorsed was good and had a good fit with the sample used, their interests, etc. The idea that the product used in the manipulations video did not had a good fit with the participants can be explained by the fact that people scored higher than the scale midpoint on their frequency of purchase of beauty products (M = 3.61, SD = 0.91). Meaning that even though in general they frequently buy beauty products, the participants had low intentions to buy the one endorsed by the influencer.
Because of the finding mentioned above concerning the frequency of purchase of beauty products (such as the product endorsed by the influencer Camila Rodrigues in the present study videos), a correlation between the purchase intentions and the frequency of purchase of these products was done. A non-significant result was found, p = .158, meaning that there is no association between how frequently the participants purchase beauty products and their purchase intentions in this study.
The results of this study generated more questions rather than answers to the ones proposed initially. However, this still provides scientific insights. With this study, it is possible to understand that maybe previously known concepts such as involvement,
entertainment education, source characteristics cannot be applied to these new medias and endorsers types, or to be able to test these variables, a pre-existing knowledge on the influencer is required or a better context should be exposed before the data collection. Regardless, it is clear that Instagram and influencers have a big impact in the present economy and social relationships, which itself is already a strong argument on why future research should understand better the processes that involve their effectivity on their multiple goal.
Reference List:
Alexander P. Schouten, Loes Janssen & Maegan Verspaget (2019): Celebrity vs. Influencer endorsements in advertising: the role of identification, credibility, and Product- Endorser fit, International Journal of Advertising, DOI:
10.1080/02650487.2019.1634898
Aguiar, T. D., & Reijmersdal, E. van. (2018). Influencer marketing. Amsterdam: SWOCC. Belch, G., & Belch, M. (1995). Introduction to advertising and promotion : an integrated
marketing communications perspective (3rd ed). Chicago [etc: Irwin.
Butcher, L., Phau, I., & Teah, M. (2016). Brand prominence in luxury consumption: Will emotional value adjudicate our longing for status? J Brand Manag, 23(6), 701–715. https://doi.org/10.1057/s41262-016-0010-8
Chung, S., & Cho, H. (2017). Fostering Parasocial Relationships with Celebrities on Social Media: Implications for Celebrity Endorsement. Psychology & Marketing, 34(4), 481–495. doi: 10.1002/mar.21001
Clement, J. (2019, November 20). Global Instagram user age & gender distribution 2019. Retrieved from https://www.statista.com/statistics/248769/age-distribution-of-worldwide-instagram-users/.
Dhanesh, G. S., & Duthler, G. (2019). Relationship management through social media influencers: Effects of followers’ awareness of paid endorsement. Public Relations
Review, 45(3), 101765. doi: 10.1016/j.pubrev.2019.03.002
Djafarova, E., & Rushworth, C. (2017). Exploring the credibility of online celebrities Instagram profiles in influencing the purchase decisions of young female users. Computers in Human Behavior, 68, 1–7. doi: 10.1016/j.chb.2016.11.009
Dodds, W., Monroe, K., & Grewal, D. (1991). Effects of Price, Brand, and Store Information on Buyers’ Product Evaluations. Journal of Marketing Research, 28(3), 307–319. https://doi.org/10.2307/3172866
Fatiha Rahmalina Fitri. (2018). The influence of celebrity endorsement in social media instagram on attitude towards brand and brand image to customer purchase intention.
Jurnal Akuntansi, Manajemen Dan Ekonomi, 20(2), 7–17. Retrieved from
https://doaj.org/article/02c150190b9e441f92484c7299f99b05
Fisher, H. D., Magee, S., & Mohammed-Baksh, S. (2015). Do They Care? An Experiment Exploring Millennials Perception of Source Credibility in Radio Broadcast News.
Journal of Radio & Audio Media, 22(2), 304–324. doi:
10.1080/19376529.2015.1083154
Giles, D. C. (2002). Parasocial Interaction: A Review of the Literature and a Model for Future Research. Media Psychology, 4(3), 279–305. doi:
10.1207/s1532785xmep0403_04
Hoffner, C. (1996). Childrens wishful identification and parasocial interaction with favorite television characters. Journal of Broadcasting & Electronic Media, 40(3), 389–402. doi: 10.1080/08838159609364360
Jamil, R. A., & Hassan, S. R. U. (2014). Influence of celebrity endorsement on consumer purchase intention for existing products: a comparative study. Journal of Management Info, 1(4), 1–8. doi: 10.31580/jmi.v4i1.18
Lafferty, B. A., & Goldsmith, R. E. (1999). Corporate Credibility’s Role in Consumers’ Attitudes and Purchase Intentions When a High versus a Low Credibility Endorser Is Used in the Ad. Journal of Business Research, 44(2), 109–116. doi: 10.1016/s0148-2963(98)00002-2
Persuasive Effects of Entertainment-Education Messages. Communication Theory,
18(3), 407–425. doi: 10.1111/j.1468-2885.2008.00328.x
Moyer-Gusé, E., Jain, P., & Chung, A. H. (2012). Reinforcement or Reactance? Examining the Effect of an Explicit Persuasive Appeal Following an Entertainment-Education Narrative. Journal of Communication, 62(6), 1010–1027. doi:
10.1111/j.1460-2466.2012.01680.x
Ohanian, R. (1990). Construction and Validation of a Scale to Measure Celebrity Endorsers' Perceived Expertise, Trustworthiness, and Attractiveness. Journal of Advertising, 19(3), pp.39-52.
Rubin, A. M., & Perse, E. M. (1987). Audience activity and soap opera involvement: A uses and effects investigation. Human Communication Research, 14, 246–268.
Schouten, A. P., Janssen, L., & Verspaget, M. (2019). Celebrity vs. Influencer endorsements in advertising: the role of identification, credibility, and Product-Endorser
fit.International Journal of Advertising, 1–24. doi: 10.1080/02650487.2019.1634898
Tracktest.eu. (n.d.). English levels CEFR description- A1,A2,B1,B2,C1,C2. Retrieved from https://tracktest.eu/english-levels-cefr/.
Appendix:
Survey English level
What is your English knowledge level? • No knowledge
• Beginner (A1)
• Elementary English (A2) • Intermediate English (B1) • Upper Intermediate English (B2) • Advanced English (C1)
• Proficiency English (C2)
Biological sex
What biological sex do you identify with most? • Male
• Female
Age
How old are you?
Fill out your age in number of years.
About influencer Next, you will watch a short video of an Instagram story from an influencer.
The name of the influencer you will see is Camila Rodrigues.
ATTENTION:
For the next step, I would like to ask you to watch a short video with sound ON. After that, answer a few questions about the video.
Do not forget to have the SOUND ON, please make sure you have your volume turned on. It is extremely important that you listen to the video's audio to follow with the questionnaire.
Make sure you pay attention to the person in the video and to this person's characteristics. Attention check
What is the product that Camila Rodrigues is endorsing? • Hair Oil from The Ordinary
• Vitamin C serum from The Ordinary
Attract scale
After watching this Instagram's influencer video, please rate the influencer on the following characteristics:
1 2 3 4 5 6 7
Unattractive • Attractive Not classy • Classy
Ugly • Beautiful Plain • Elegant Not sexy • Sexy
Trustworthy scale
After watching this Instagram's influencer video, please rate the influencer on the following characteristics: 1 2 3 4 5 6 7 Undependable • Dependable Dishonest • Honest Unreliable • Reliable Insincere • Sincere Untrustworthy • Trustworthy
PSI questions
Please state to what extent do you agree or disagree with the following statements concerning the video you watched:
Strongly agree Agree Somewhat agree Neither agree nor disagree Somewhat disagree Disagree Strongly disagree "Camila Rodrigues makes me feel comfortable, as if I am with a friend." (1) • "I would like to have a friendly chat with Camila Rodrigues." (2) • "If Camila Rodrigues were not a celebrity, we would have been good friends." (3) •
WI questions
Please state to what extent do you agree or disagree with the following statements concerning the video you watched:
Strongly agree Agree Somewhat agree Neither agree nor disagree Somewhat disagree Disagree Strongly disagree "I would like to do the kind of things Camila Rodrigues does on her day-by-day." (1) • "Camila Rodrigues is the sort of person I want to be like myself." (2) • "I wish I could be more like Camila Rodrigues." (3) •
Purchase Int quest
Please state to what extent do you agree or disagree with the following statements concerning the video you watched:
Strongly agree Agree Somewhat agree Neither agree nor disagree Somewhat disagree Disagree Strongly disagree "I am likely to buy this product" (1) • "I am willing to buy this product" (2) • "I would actively seek out this product in a store in order to purchase it" (3) •
Freq of beauty prod How often do you buy beauty products? • Always • Often • Sometimes • Rarely • Never
Which social media Which of the following social media do you use? • Facebook • Instagram • YouTube • Tumblr • WhatsApp • LinkedIn • Twitter
