Disclosing or Disguising
Influencer Marketing on Instagram?
The Impact of Disclosures, Cues and Influencer Types on Users’ Ad Recognition and Responses towards the Persuasive Message, the Influencer and the Advertised Brand
Research Master’s Thesis Céline Marie Müller (11802502) Supervisor: Dr. Sophie C. Boerman
June 28, 2019
Graduate School of Communication Research Master Communication Science
Abstract
Influencer marketing blurs the lines between editorial content and commercial Instagram posts, creating difficulties for consumers to recognize advertising. Although regulatory parties recommend the use of unambiguous disclosures, ‘Instagrammers’ frequently refrain from clearly disclosing their brand partnerships and merely include vague cues in their posts, such as brand tags. With a 4 (disclosure: ‘Paid partnership’-label, #ad, #paidad, none) x 4 (cue: brand tag in picture, #brand, @brand, brand mention in caption) within-subjects eye-tracking experiment (N = 60), this study first aimed to clarify whether users recognize advertising and which elements they use to do so. For further insights, we conducted an online experiment (N = 433) with a 4 (disclosure/cue type: ‘Paid partnership’-label, #paidad, brand tag in picture, no disclosure) x 2 (influencer type: macro-influencer, nano-influencer) between-subjects design. This second study intended to detect whether influencer types function as a cue for ad recognition and, whether the elements found to increase ad
recognition in the first study affect users’ reactions towards the post, the influencer and the advertised brand. The findings demonstrate that both disclosures and cues increase users’ ad recognition. While there is merely tentative support for a moderation effect of Instagrammer type on this effect, posts by macro-influencers compared to nano-influencers are more likely to be recognized as an advertisement. Furthermore, increased recognition of advertising due to disclosures or cues leads to increased brand recall and to higher skepticism towards the post, with the latter resulting in lower influencer trustworthiness, brand attitudes and purchase intention.
Introduction
With 895 million monthly-active users worldwide (Kemp, 2019), Instagram has become highly relevant for brands to reach young audiences, who are difficult to approach via traditional advertising (Domingues Aguiar & Van Reijmersdal, 2018). The social media platform facilitates targeted access to consumers via influencers – Instagram users with an established followership and trustworthiness (De Veirman, Cauberghe, & Hudders, 2017). Influencers “typically receive products for free […] or are being paid […] to include and recommend brands on their social media profiles and, in this way, shape their followers’ opinions” (De Veirman & Hudders, 2019, p. 3). This marketing strategy – influencer marketing – seems less obtrusive than traditional advertising since commercial content is incorporated into editorial posts (Wojdynski & Evans, 2016). Therefore, users may be fooled into thinking that the influencers’ posts reflect personal opinions instead of being attributable to paid partnerships with brands (Evans, Phua, Lim, & Jun, 2017a; Coursaris, Van Osch, & Kourganoff, 2018). Hereby, advertisers overcome consumers’ resistance towards persuasive messages, as they may not recognize this type of advertising (Boerman & Van Reijmersdal, 2016).
Consequently, regulatory parties, such as The Federal Trade Commission (FTC), recommend sponsorship disclosures, which clearly indicate posts’ commercial purpose. The few studies that explored Instagram users’ responses to disclosures focused on hashtags, such as #ad, or Instagram’s platform-specific ‘Paid partnership with [brand]’-label and found that explicit disclosures increase consumers’ ad recognition (Evans et al., 2017a; Coursaris et al., 2018). However, merely 25% of Instagram influencers disclose commercial posts with clear disclosures, that comply with FTC regulations (“The state of disclosure,” 2018). Instead, brand mentions and (hash-)tags in posts’ captions or brand tags in the pictures often indicate influencers’ collaboration with brands (InfluencerDB, 2017). Although these cues do not clearly disclose ads, consumers may have learned to use them to identify advertising. This
thesis unprecedently adapts to reality by considering both the effectiveness of explicit disclosures and ambiguous, but widely-used, cues on users’ ad recognition. With an eye-tracking experiment, this study first aims to clarify whether Instagram users recognize influencer marketing and which disclosures or cues they use to do so. Eye-tracking is highly beneficial in this context, since it reveals which elements attract users’ attention and, thus, suggest which visual information they use to identify influencer marketing.
Besides helpful elements in the post, Coursaris et al. (2018) found that consumers are more likely to recognize ads by ‘Instagrammers’ with many followers, supposedly because they are aware that celebrities or popular influencers are often paid to endorse a product or brand. Hence, there is reason to believe that influencer types impose boundary conditions on disclosures’ necessity, depending on whether users expect influencer marketing by an ‘Instagrammer’, regardless of any disclosures. Such boundaries may be detected by comparing an increasingly demanded small-scale group of ‘Instagrammers’, called nano-influencers, who have not been considered in academia before, to their exact opposite, internationally known macro-influencers.
Further, ad recognition – whether activated by disclosures or other cues – is known to affect consumers’ responses towards the persuasive message, the influencer and the
advertised brand (i.e. De Veirman & Hudders, 2019; Evans et al., 2017a). Thus, the second aim of this study is to detect whether influencer types function as a cue for ad recognition and whether the elements found to be helpful in the first study – be it explicit disclosures or vague cues – entail the above-mentioned consequences on users’ reactions.
Concluding, by conducting an eye-tracking study and a subsequent online experiment, the present thesis intends to answer the following overall research question:
To what extent do elements in an Instagram post and the type of influencer affect users’ ad recognition and further cognitive and evaluative responses towards the persuasive message, the influencer and the advertised brand?
Ad Recognition in the Context of Influencer Marketing on Instagram
According to the Persuasion Knowledge Model (PKM), consumers develop so-called persuasion knowledge (PK), which refers to their understanding of marketers’ goals and tactics in persuasive attempts, their evaluation of these tactics and their beliefs about effective coping mechanisms in such situations (Friestad & Wright, 1994). This knowledge is
constituted of a cognitive and an affective dimension, respectively conceptual and attitudinal PK (Rozendaal, Lapierre, Van Reijmersdal, & Buijzen, 2011; Boerman, Van Reijmersdal & Neijens, 2012). Ad recognition, a fundamental dimension of conceptual PK, is of particular importance in this study. Only when consumers recognize persuasive attempts, their PK is activated to construe, assess and react to such attempts (Boerman, Van Reijmersdal, & Neijens, 2015; Muñoz-Leiva, Hernández-Méndez, & Gómez-Carmona, 2019).
Since ‘Instagrammers’ frequently share commercial content without adequate disclosures, users may face difficulties in discerning the persuasive intent behind such posts (Evans et al., 2017a; Coursaris et al., 2018). Simultaneously, due to the growing amount of influencer marketing on Instagram (De Veirman & Hudders, 2019), avid users may have developed sufficient PK to identify influencer marketing without needing a disclosure – for instance, based on a post’s strong emphasis on a product. Indeed, studies in the social media context found ad recognition rates as high as 97% for Instagram ads (Johnson, Potocki, & Veldhuis, 2019), 82% for native Facebook ads (Jung & Heo, 2019) and 84% for sponsored user-generated tweets (Kim & Song, 2018). However, in contrast to these studies, the present experiment takes into account ambiguous cues next to clear disclosures, which may lower consumers’ ad recognition. Hence, this thesis first seeks to answer the question:
RQ1: Do users recognize influencer marketing on Instagram?
Fogg’s (2003) prominence-interpretation theory suggests that individuals need to notice and understand an object before they can evaluate it. The first component, prominence, refers to a person’s likelihood of noticing and paying attention to an element. In the context
of the present study, eye-tracking allows for an accurate measurement of individuals’ attention to specific elements in an Instagram post and, thus, grants insights into the disclosures or cues that attract the most visual attention (Boerman et al., 2015).
There are several ways to disclose influencer marketing. Instagram introduced the standardized disclosure ‘Paid partnership with [brand]’ in 2017, which is located in-between the name of the ‘Instagrammer’ and the post. This appears as a more prominent position than hashtags such as #sponsored, which are placed underneath the post and may go unnoticed alongside further text in the caption. Yet, the FTC assumes that Instagram’s standardized disclosure is insufficient because it may not attract users’ attention (FTC 2017a, 2017b). Hence, the present study analyses this disclosure to assess whether it can effectively increase ad recognition, or whether the FTC is rightly suggesting that it is inadequate.
Furthermore, ‘Instagrammers’ commonly disclose influencer marketing by adding hashtags to their captions. The most-frequently used hashtag to declare commercial posts is #ad (Hellenkemper, 2017). However, #ad is not the most effective disclosure hashtag in terms of increasing ad recognition. Evans et al. (2017a) found #paidad to be the most powerful hashtag to activate persuasion knowledge. Due to #ad’s practical relevance and #paidad’s reported efficiency, both these hashtags are examined in this study.
Besides these unambiguous disclosures, ‘Instagrammers’ often rely on vague cues that do not clearly reveal a post’s commercial purpose (InfluencerDB, 2017). Due to the frequent application of such cues, consumers may have learned to use them to identify influencer marketing. These cues mostly contain the brand name and, thus, immediately reveal the advertised brand. Brands may be mentioned (brand), tagged (@brand) or hashtagged (#brand) in the caption. Additionally, brands may be tagged in the post itself. Despite their frequent appearance, such cues have been neglected in prior research. The present study considers these four cues to examine their effectivity in raising ad recognition.
Brand (hash-)tags are colored blue in the caption or displayed in a black box in the picture and redirect the user to the branded hashtag or the brand account upon clicking on them. In contrast, simple mentions appear exactly like back, regular text in the caption and do not contain a link to the brand’s Instagram page or the like. Hence, this cue type may be characterized as the least salient and dynamic cue. Importantly, studies have found that elements’ position, color and size affect viewers’ visual attention (e.g. Wojdynski et al., 2017; Wojdynski & Evans, 2016; Boerman et al., 2015). This suggests that the plain brand mention may attract less attention than the (hash-)tags.
Results on the effect of attention on ad recognition are mixed. While Boerman and colleagues (2015) observed that visual attention to a disclosure mediates the effect of disclosure type on ad recognition, other studies found that merely seeing a disclosure does not guarantee increased ad recognition (Wojdynski et al., 2017; Smink, Van Reijmersdal, & Boerman, 2017). Due to inconsistent findings, the second research question asks:
RQ2: Which disclosures or cues in an Instagram post do users attend to most and do these elements help them recognize influencer marketing?
Using eye-tracking allows to literally see whether specific disclosures or cues catch users’ attention longer than others. The gathered data are particularly useful since they grant insights into what consumers pay attention to as they scroll through an Instagram feed. By directly observing eye movements, we can unobtrusively measure, which disclosure or cue attracts attention most effectively (Boerman et al., 2015; Josephson & Miller, 2015). Method – Study 1: Eye-Tracking Experiment
Design and Participants. An eye-tracking study with a 4 (disclosure: standardized disclosure, #ad, #paidad, none) x 4 (cue: brand tag in picture, #brand, @brand, brand mention in caption) within-subjects design was conducted in the university’s lab from April 23 to May 1, 2019. A total of 72 participants were recruited through the university-internal website for participant recruitment, as well as flyers that were spread in university. Subjects were
informed that they were participating in an eye-tracking study on people's reactions towards social media posts. Of this sample, 12 participants were excluded because they did not have an Instagram account, were not proficient in English or due to defective records of their eye movements, leading to a final sample of 60. The majority of participants were female (78.33%), with an average age of 22 years (M = 22.35, SD = 4.20). More than half of them indicated high school as their highest completed level of education (56.67%), followed by 31.67%, who completed at least a Bachelor’s. This reflects the sample’s student character. Most participants use Instagram multiple times a day (76.67%).
Procedure. Upon arrival, participants read an introductory text and signed informed consent before being asked to sit comfortably behind a 22-inch screen with the eye-tracker. The screen was placed approximately 21 to 28 inches from the participant. After successful 9-point calibration, participants were exposed to a video scrolling through an Instagram feed (see Appendix I, Figure 1 and 2). Binocular eye movements were registered using the SMI RED eye-tracker with a gaze sample rate of 120 Hz per second. Participants then continued to the questionnaire, in which they indicated for each of the posts they saw in the video whether they think it contained advertising. Subsequently, participants answered questions about brand and disclosure memory as well as their PK. The questionnaire ended with control questions and demographics. Ultimately, participants were debriefed, thanked and received either 5€ or two research credits for taking part in the study.
Stimuli. The stimulus material was a 4:12 minute-long video scrolling through an Instagram feed consisting of 50 posts. Each post was shown statically for three seconds while scrolling to the next post took 2 seconds. For posts containing brand tags in the picture, the post was briefly shown without the tag, the tag was then displayed for three seconds before disappearing again and scrolling onto the next post. Half of the posts contained advertising while the other half were non-commercial fillers. All ‘Instagrammers’, with varying degrees of popularity ranging from 1,000 to 112 million followers, were real and actually posted
these pictures. However, elements such as the captions or disclosure were adjusted. Each of the 25 commercial post contained a combination of one of the above-mentioned disclosures (standardized disclosure, #paidad, #ad, no disclosure) and one of the cues (@brand, #brand, brand, pictorial brand tag). As the experimental design yields 16 conditions, some variations occurred twice because 25 manipulated posts were displayed (see Appendix I for video).
Measures.
Visual attention to the Disclosures and Cues. With the SMI BeGaze software, the
recorded eye-tracking data were prepared and exported. For all disclosures and cues, individual areas of interest (AOIs) were created for the exact period they were visible in the video (see Appendix I, Figure 3). Visual attention was measured with fixation time in milliseconds within each AOI. A fixation was measured whenever the eyes stayed at a point for at least 80 milliseconds. Fixation time is considered a valid indicator of attention that reflects participant’s processing depth (King, Bol, Cummins, & John, 2019).
Ad recognition. Ad recognition was measured by showing participants an overview
of all pictures they saw in the video and asking them to select those that they thought
contained advertising (0 = No, 1 = Yes). Overall, participants correctly recognized two thirds (68.60%) of the posts that contained advertising. Furthermore, participants indicated how they knew that a post contained advertising in an open text field (Wojdynski & Evans, 2016).
Control variables. Besides participants’ age in years and gender, frequency of
Instagram use was measured by asking participants to indicate how often they use Instagram (1 = Never, 2 = Yearly, 3 = Monthly, 4 = Weekly, 5 = Approximately once a day, 6 = Multiple times a day; M = 5.47, SD = 1.19). Additionally, participants were asked about their opinion on Instagram, measured on a 7-point semantic differential scale composed of 3 items
(Dislike-Like, Negative-Positive, Bad-Good, Eigenvalue = 2.42, explained variance = 80.60%, = .88, M = 5.14, SD = 1.39) and whether they already knew that some Instagram posts contain advertising before participating in this study (0 = No, 1 = Yes, 90.00% said yes).
Results and Implications for Study 2
To answer the first research question, descriptive statistics revealed how successful participants were in identifying influencer marketing. The most-recognized post featured a combination of the standardized and a brand tag in the picture (90.00% recognized ad; see Appendix II, Table 1 for an overview of all posts’ ad recognition). On average, participants identified 17.15 (SD = 4.44) of the 25 commercial posts and, thus, recognized two in three ads. Simultaneously, participants erroneously classified an average of 4.42 posts (SD = 3.04) of the 25 filler posts as advertising, which equates to 17.68% of the non-commercial posts.
The second research question asked which elements in a post attract users’ attention and whether these elements help them recognize ads. For answering this question, fixation times to the AOIs were grouped based on disclosure or cue type. Pairwise comparisons among the attention spent on these groups revealed that the highest scoring elements (see Table 1 below) with respect to visual attention – brand tags in pictures and standardized disclosures – barely differed from each other, but caught significantly more attention than all other disclosures or cues (see Appendix II, Table 2 for pairwise comparisons). These two elements were followed by brand hashtags and brand tags in the caption, the disclosure #paidad and the mere brand mention in the caption. Ranking last, participants paid significantly less attention to the disclosure #ad compared to all other disclosures. Table 1
Fixation Time per Disclosure Type in Milliseconds
Disclosure/Cue M SD
Brand tag in picture 269.19a 271.50
Standardized disclosures 244.38a 176.03
#brand in caption 151.03b 121.42
@brand in caption 109.15bc 103.65
#paidad in caption 99.83bc 92.40
Brand mention in caption 85.26c 80.42
#ad in caption 44.96d 63.46
Both hashtag disclosures #paidad and #ad received slightly more attention when placed in the beginning of the caption compared to the end although these differences were not significant (#paidad: MStart = 102.22, SDStart = 129.09 vs. MEnd = 97.43, SDEnd = 119.16, p = 1.000; #ad: MStart = 58.53, SDStart = 101.58 vs. MEnd =38.18, SDStart = 70.58, p = .963).
To clarify which disclosures and cues helped users’ to recognize ads, participants’ open answers on how they knew that posts contained advertising were evaluated.
Corresponding to the visual attention results, the standardized disclosure and brand tag in picture were mentioned most often when participants were asked how they recognized ads, both by around 45% of the sample. However, in contrast to the findings on visual attention, a third indicated the disclosure #ad as helpful. Brand mentions or tags were cited by almost every fourth person, followed by #paidad, being mentioned by one in five participants. Interestingly, 11.67% of the sample explained that they did not recognize influencer
marketing based on disclosures or brand mentions, but figured it was advertising based on the setup of the picture or because it was posted by an influencer.
Concluding, the results indicate that today’s Instagram users have developed
conceptual PK to an extent that allows them to identify more than two thirds of commercial posts. This finding is slightly lower than recent studies that found high ad recognition for advertising on social media (Johnson et al., 2019; Jung & Heo, 2019; Kim & Song, 2018). The present study’s findings suggest that Instagram users, though already able to recognize the majority of ads, still lack sufficient PK to accurately distinguish between non-commercial content and influencer marketing at all times.
Furthermore, results demonstrate that subjects bring more attention to hashtags places in the beginning of the caption compared to the end and that brand tags in the picture and the standardized disclosure catch users’ attention the longest. This is in line with previously mentioned studies, which found that a prominent position, color and size increase visual attention (e.g. Wojdynski et al., 2017).
Moreover, Instagram’s standardized disclosure and brand tag in picture did not only catch significantly more attention than the remaining elements, but were also mentioned most often as a reason why participants recognized influencer marketing. Hence, a brand tag in the picture – a cue that is not classified as a sufficient disclosure by regulatory parties such as the FTC – proved to be highly relevant in terms of visual attention and users’ recognition of ads. Due to the within-subject design, in which each post contained both a disclosure and a cue, it is difficult to separate these elements and indicate which particular disclosure or cue achieved the highest ad recognition. Hence, based on the visual attention paid to the
individual disclosures or cues, specific elements were selected for the subsequent online experiment. Therefore, the second study is essential to demonstrate clear effects of particular elements on consumers’ ad recognition and further responses.
Effects of Different Disclosures or Cues on Ad Recognition
Based on the findings of the first study, an online experiment was conducted to elucidate to what extent different disclosures or cues as well as influencer types affect users’ PK and responses towards the brand and the influencer. In the eye-tracking experiment, Instagram’s standardized disclosure was found to gain most visual attention compared to the other explicit disclosures. The disclosure ‘#paidad’ proved to be the second-most visually attended disclosure, particularly when placed at the beginning of a post’s caption. #paidad’s importance was also determined in a study by Evans and colleagues (2017a). Although not an explicit disclosure, the eye-tracking experiment revealed that brand tags in picture captured users’ visual attention the longest, exceeding all explicit disclosures and other cues.
Therefore, Instagram’s standardized disclosure, #paidad in the beginning of the caption, brand tag in picture (hereafter referred to as ‘brand tag’ for the sake of simplicity) and the absence of a disclosure are compared in the second study.
Figure 1. Example post including the disclosures or cues tested in study 2
Regarding different disclosure types, previous studies indicated that disclosure positions vary in terms of their effectiveness on recognition of advertising. Boerman, Van Reijmersdal and Neijens (2014) found that disclosures placed above the content best activate ad recognition. Additionally, a study by Wojdynski and colleagues (2017) demonstrated that transparent and informative disclosures revealing both the commercial purpose and the advertised brand foster ad recognition. Combined with the eye-tracking results, these findings suggest that Instagram’s standardized disclosure is more effective in increasing ad
On the one hand, Evans and colleagues (2017a) showed that, compared to other hashtags, #paidad is most effective in stimulating ad recognition. On the other hand, brand tags caught participant’s attention significantly more than #paidad in the preceding eye-tracking study. Additionally, middle-positioned and prominent disclosures were previously identified as stimulators of ad recognition (Wojdynski et al., 2017), suggesting that brand tags may exert a stronger effect on ad recognition than #paidad. Still, #paidad is expected to result in higher ad recognition than a brand tag in the present study. Although the tag distinctly displays the brand name, it does not communicate clear information about the relationship between this brand and the influencer. Therefore, these tags may exert a lower influence on ad recognition compared to the two disclosures. Summarizing, Instagram’s standardized disclosure is expected to result in the highest recognition of advertising, followed by #paidad. A brand tag is hypothesized to lead to a lower ad recognition than #paidad, but to a higher ad recognition than no disclosure:
H1: The type of disclosure or cue influences users’ ad recognition, with ad
recognition being the highest for Instagram’s standardized disclosure, followed by #paidad, which is then followed by brand tag. Ad recognition is expected to be lowest in absence of a disclosure.
The Moderating Effect of Influencer Type on Ad Recognition
Based on their amount of followers, influencers may be classified into different types. Influencer marketing practiced by nano-influencers, users with less than 1000 followers, is a relatively new phenomenon, that is increasingly pursued by brands due to the characteristics of these ‘Instagrammers’. Nano-influencers have a small and specifically-niched follower base and may be perceived as more relatable and trustworthy by the ordinary user as they are close to their followers and actively engage with them (Cohn, 2019; Talbot, 2018). Micro-influencers have up to 10,000 followers and, thus, do stand out against the average user (Hatton, 2018). Meso-influencers reach up to a million followers and often have national
visibility (Verheye, 2017). Macro-influencers are internationally-known ‘Instagrammers’ with over one million followers who often rose to fame as models, bloggers, actors or athletes (Hatton, 2018; Pedroni, 2016). The present study focuses on nano-influencers, due to their novel importance (Cohn, 2019; Talbot, 2018) and on macro-influencers, due to their mere size and relevance. This is practical in detecting potential boundary conditions of disclosures’ necessity by contrasting two tremendously different influencer types: internationally-popular and unknown ‘Instagrammers’.
Research has shown that users are aware that Instagram is used for commercial purposes (Djafarova & Trofimenko, 2018; Chen, 2018) and, thus, may have already developed PK about influencers’ persuasive techniques. Comparable to celebrity
endorsement in traditional advertising, users may have established that ‘Instagrammers’ with a large follower base are more likely to be paid for implementing products or brands in their posts (Coursaris et al., 2018). Besides number of followers, the type and title of an Instagram account also provides an insight into the status of the influencer. More specifically, verified accounts, which mostly belong to well-known ‘Instagrammers’, feature a blue check that appears next to an account's name in search and on the profile. It means Instagram has confirmed that an account is the authentic presence of the public figure, celebrity or global brand it represents. Furthermore, famous ‘Instagrammers’ often include a label (i.e. ‘Public figure’) below their name on their account overview, indicating their celebrity status.
Supposing that an influencer’s prominence indeed functions as a cue that elicits ad recognition, boundary conditions may apply to the essentiality of sponsorship disclosures. More specifically, even in the absence of disclosures or cues (hereafter generically referred to as ‘disclosure cues’ for the sake of simplicity), users may suspect influencer marketing by macro-influencers, thus, rendering disclosure cues less effective. Inversely, nano-influencers are more similar to the average user and, hence, may be considered less likely to advertise a brand (Coursaris et al., 2018). Therefore, when disclosure cues are absent, users might
believe that nano-influencers’ brand endorsements in posts reflect personal opinions (Evans et al., 2017a). Henceforth, an interaction effect of disclosure cue and influencer type on conceptual PK is proposed:
H2: The effect of disclosure cues, compared to their absence, on conceptual persuasion knowledge is moderated by the type of influencer: The disclosure cue has a stronger effect on conceptual persuasion knowledge when the post is sent by a nano-influencer, compared to a macro-influencer.
Consequences of Disclosure Cues and Ad Recognition
Cognitive consequences. Research suggests that sponsorship disclosures positively influence consumers’ brand recall (Boerman & Van Reijmersdal, 2016). The limited capacity model proposes that people have a limited capacity to encode, store and retrieve information (Lang, 2000). However, a sponsorship disclosure draws attention to the commercial purpose of a message and, thus, the likelihood that an individual notices this disclosure is increased (Boerman et al., 2014). Hence, the disclosure functions as a cue, which may increase consumers’ likelihood to remember the advertised brand. Moreover, the standardized disclosure and the brand tag display the brand name, thus, increase exposure to the name of the brand, which may make it easier for consumers to recall the brand. In line with this theoretical reasoning, studies have found that disclosures indirectly affect brand memory via ad recognition (Smink et al., 2017; Boerman et al., 2015). Therefore, the third hypothesis is: H3: A disclosure cue in an Instagram post, compared to the absence of it, increases conceptual persuasion knowledge, which leads to a higher brand recall.
Affective consequences. In addition to the proposed cognitive route affecting brand recall, this study proposes a second, evaluative route involving attitudinal PK, influencer trustworthiness, brand attitudes and purchase intention.
Besides its cognitive dimension, PK encompasses an affective side, referred to as attitudinal PK, which concerns consumers’ critical feelings towards a persuasive attempt. It is
believed that consumers’ activation of attitudinal PK is contingent on their ad recognition (Boerman, Van Reijmersdal, Rozendaal, & Dima, 2018; Rozendaal et al. 2011). Reactance theory (Brehm, 1966) assumes that individuals tend to fear that they are being manipulated when they are subject to a persuasive attempt. Thus, when they recognize persuasive intent, they are likely to develop reactance to maintain their freedom. Correspondingly, in their study on Instagram, De Veirman and Hudders (2019) found that sponsorship disclosures increased ad recognition, which consequently stimulated ad skepticism. Ad skepticism refers to “the tendency towards disbelief of sponsored content” (Boerman et al., 2018, p. 675) and plays an important role in this study because of influencer marketing’s covert nature, which may cause users to feel mislead. Due to this impression of deception, consumers tend to react to commercial content with increased skepticism (Boerman et al., 2014).
Furthermore, the PKM proposes that, when consumers realize the persuasive intent behind a message, a change of meaning may occur, which fundamentally alters consumers’ responses and feelings towards the message (Friestad & Wright, 1994). Previous studies found that such critical feelings towards an ad also spill over to the targets’ responses towards the sender’s credibility (De Veirman & Hudders, 2019), the brand (Boerman 2012) and their intention to purchase the advertised product (Evans, Wojdynski, & Hoy, 2017b). Therefore, potential effects of persuasion knowledge on individuals’ perceived trustworthiness of the influencer, brand attitudes and purchase intentions are discussed hereinafter.
Carr and Hayes (2014) detected that, in the context of online blogs, explicit sponsorship disclosures enhance the bloggers’ trustworthiness, showing that consumers appreciate their honesty about relationships with brands. In contrast, De Veirman and Hudders (2019) revealed that sponsorship disclosures on Instagram lower the influencer’s credibility via conceptual and attitudinal PK. This is supported by the PKM, which predicts that increased PK negatively affects consumer responses (Friestad & Wright, 1994). Thus, when users recognize the commercial purpose behind a post and realize the influencer’s true
motives, they become more critical towards the post, which causes them to perceive the influencer as less trustworthy. Similarly, various studies found that the activation of conceptual and attitudinal PK through disclosures affects consumers’ attitudes toward the advertised brand negatively (Smink et al., 2017; Van Reijmersdal et al., 2016; Meijers, Van Reijmersdal, & Krafczyk, 2018; Boerman et al., 2012, 2014, 2015). Finally, induced by reactance and a change of meaning, a spill over to behavioral reactions may reduce the target’s intention to purchase the advertised product (Evans et al., 2017a; Lu, Chang, & Chang; 2014; Tessitore & Geuens, 2013). Summarizing, the present study suggests that disclosure cues will stimulate an evaluative process, in which consumers’ perceptions of the influencers’ trustworthiness, their brand attitudes and purchase intentions are affected via conceptual and attitudinal PK.
H4: A disclosure cue in an Instagram post, compared to its absence, increases conceptual persuasion knowledge, which elicits a) higher attitudinal persuasion knowledge, which consequently leads to b) lower trustworthiness of the influencer, c) lower brand attitudes and d) lower purchase intention.
Based on the above-elaborated expectations, Figure 2 presents the conceptual model that forms the basis of the second study.
Figure 2. Conceptual Model of the Online Experiment (Study 2). Method – Study 2: Online Experiment
Design and Participants. To test these hypotheses, we conducted an online
(influencer type: macro, nano) between-subjects design between May 30 and June 3, 2019. A total of 460 participants were recruited through the crowdsourcing platform Prolific based on whether they fulfilled the criteria of having an Instagram account, being of British nationality and being fluent in English. The final sample consisted of 433 people as 27 participants who either failed both attention checks or indicated they never use Instagram were excluded (MAge = 31.67, SDAge = 10.64, 74.83% female, 39.03% completed Bachelor’s degree). Half of the participants use Instagram multiple times a day (49.88%).
Procedure. Participants were informed that they were participating in an online experiment on people's reactions towards social media posts, granted informed consent and were randomly assigned to one of the eight conditions. Initially, participants were exposed to an overview of an Instagram account, in which influencer type was manipulated, followed by a post of that ‘Instagrammer’, in which disclosure cue type was manipulated. Participants had to view both the overview and the post for 10 seconds before being able to switch to the next page. The ensuing questionnaire entailed questions about participants’ conceptual and attitudinal PK, responses to the advertised brand as well as brand and disclosure memory, followed by control questions, manipulation checks and demographic questions. Upon completion, participants were debriefed, thanked and received a compensation (£1).
Stimuli. The stimulus materials consisted of an overview of the account of an
‘Instagrammer’ and one post by that Instagrammer. In both conditions, the post contained the same picture taken from a Dutch influencer’s account (@dee), which was selected because it could be used for both conditions and because it displayed a brand. More specifically, the picture showed a blonde woman whose face was not identifiable since she turned her back to the camera, trailing a suitcase by the brand Suit Suit. In both conditions, the alleged accounts of two real ‘Instagrammers’ were displayed who had the same biography and highlights on their profile but were manipulated by number of followers, account type and ‘Instagrammer’ title. The macro-influencer condition showed the profile of model and actress Doutzen Kroes
with six million followers, a verified account, meaning it featured a blue check that communicates her authenticity and the title ‘public figure’ underneath Kroes’ name. The nano-influencer condition displayed the profile of ‘Instagrammer’ Melissa Bell with 715 followers, a non-verified account, meaning no blue check was visible, and the title ‘blogger’ underneath Bell’s name. Participants were first exposed to an overview of one of the accounts and instructed to take their time to look at it since they would be asked questions about it afterwards. Above the account overview, a text introduced each ‘Instagrammer’ differently, drawing participants’ attention to the number of followers and whether the account is verified or not.
Subsequent to the overview, participants were exposed to an Instagram post by the respective influencer, equal with regard to the post’s caption and the amount of likes in both conditions. The disclosure was manipulated by either placing the standardized label ‘Paid partnership with suitsuit_’ above the picture, mentioning #paidad in the caption, tagging the brand in the picture or not disclosing the advertisement (see Figure 3 for account overview of macro- and nano-influencer and Appendix I, Figure 4 to 11 for stimuli of all conditions).
Measures. See Table 2 for a detailed overview of all measures. Table 2
Measures: items and descriptive statistics
Items Measurement Source Descriptives
Ad recognition
Please indicate the extent to which you agree or disagree with the following statements.
The post I just saw contained advertising, The post I just saw showed or mentioned brands,
The post I just saw was commercial, The post I just saw was paid by a brand
7-point scale (1 = strongly disagree, 7 = strongly agree) Boerman et al., 2018; Van Reijmersdal et al., 2016 Eigenvalue = 3.07, explained variance = 76.69%, = .90; M = 4.04, SD = 1.59 Ad skepticism
Brands sometimes pay people on Instagram to mention or show the brand in their Instagram posts. What is your opinion about this?
I think that showing the brand Suit Suit in the Instagram post is...
dishonest/honest not trustworthy/trustworthy incredible/credible not truthful/truthful insincere/sincere 7-point semantic differential scale, reverse coded Boerman et
al., 2018 Eigenvalue = 3.96, explained variance = 79.27%,
= .93; M = 4.14, SD = 1.23
Brand recall
Do you recall seeing any brands in the Instagram post?
No
Yes, namely [text entry]
Coded as 0 when recalled incorrectly or not at all and 1, when recalled correctly Boerman et al., 2012, 2015; Smink et al., 2017 54.5% recalled correctly, M = .54, SD = .50
Items Measurement Source Descriptives Influencer trustworthiness
What is your opinion of the woman posting the Instagram post? undependable/dependable dishonest/honest, unreliable/reliable, insincere/sincere, untrustworthy/trustworthy 7-point semantic differential scale Ohanian,
1990 Eigenvalue = 3.72, explained variance = 74.35%,
= .91; M = 4.01, SD = 1.00 Brand attitudes
What is your opinion of the brand Suit Suit? bad/good, negative/positive, dislike/like, unpleasant/pleasant, unfavorable/favorable 7-point semantic differential scale Spears &
Singh, 2004 Eigenvalue =4.14, explained variance = 82.71%,
= .95; M = 4.33, SD = .87 Purchase intention
Imagine you need a suitcase.
How likely are you to buy a suitcase of the brand Suit Suit?
7-point scale (1 = very unlikely, 7 = very likely) Bergkvist & Rossiter, 2009 M = 2.85, SD = 1.51 Manipulation Check Influencer 1
Some people on Instagram have a verified account. These accounts have a blue check that appears next to an Instagram account's name in search and on the profile. It means Instagram has confirmed that an account is the authentic presence of the public figure, celebrity or global brand it represents. Do you recall whether the Instagram user you just saw had a verified account? Manipulation Check Influencer 2 How many follower do you think the Instagram user you just saw has? 0 – 1,000 1,001 – 10,000 10,001 – 50,000 50,001 – 250,000 250,001 – 1,000,000 More than 1,000,000 0 = No, the Instagram user did not have a verified account, 1 = Yes, the Instagram user had a verified account Depending on condition, recoded to 1 = correct answer, 0 = incorrect answer 7.4% failed, M = .93, SD = .26 25.5% failed, M = .74, SD = .44
Items Measurement Source Descriptives Manipulation Check Disclosure
Do you recall seeing one of the following hashtags or statements disclosing that the post contains advertising? You can only choose one option.
#ad, #paidad, #sponsored,
Paid partnership with [brand], Sponsored content,
Advertising, None of the above
Depending on condition, recoded to 1 = correct answer, 0 = incorrect answer 29.8% failed, M = .70, SD = .46 Control variables Age
What is your age (in years)? [open answer] M = 31.67,
SD = 10.64 Gender
Are you… 0 = male,
1 = female, 2 = other
74.8% female, no one chose ‘other’, M = .75,
SD = .44 Education level
What is the highest level of education
that you have completed? 1 = Less than high school, 2 = High school, 3 = Some college but no degree, 4 = Associate degree, 5 = Bachelor’s degree, 6 = Master’s degree, 7 = Doctorate Degree (PhD) Majority (39%) completed a Bachelor’s degree; M = 4.05, SD = 1.43
Frequency of Instagram use On average, how often do you use
Instagram? 1 = Never, 2 = Yearly,
3 = Monthly, 4 = Weekly, 5 = Approx. once a day, 6 = Multiple times a day Majority (49.9%) uses Instagram multiple times a day M = 5.17,
Items Measurement Source Descriptives Attitude towards Instagram
What is your opinion of the social media platform Instagram? dislike/like, negative/positive, bad/good 7-point differential semantic scale Eigenvalue = 2.63, explained variance = 87.63%, = .93; M = 5.37, SD = 1.33 Brand familiarity
Before participating in this study, did you
already know the brand Suit Suit? 0 = No, 1 = Yes 1.8% familiar with brand, M = .02,
SD = .14 Product Interest
Please indicate the extent to which you agree or disagree with the following statements.
I like buying suitcases and bags I like seeing something about suitcases and bags on social media
I am interested in suitcases and bags
7-point scale (1 = very unlikely, 7 = very likely) Eigenvalue =2.51, explained variance = 62.90%, = .80; M = 3.87, SD = 1.41 Results Manipulation Checks
For the first manipulation check on influencer type, 92.09% in the macro-influencer conditions said they did see a verified badge while 93.12% in the nano-influencer conditions indicated they did not see a verified badge. This difference was significant, χ2(1) = 314.47, p < .001. Additionally, 66.51% in the macro-influencer condition correctly remembered the ‘Instagrammer’ having more than 1 million followers while 82.11% in the nano-influencer condition accurately stated that the ‘Instagrammer’ had less than 1.000 followers. This difference was significant, χ2(5) = 370.24, p < .001.
In the no disclosure conditions 90.91%, in the brand tag conditions 90.99% and in the #paidad conditions 55.24% correctly passed the manipulation check. Of those exposed to Instagram’s standardized disclosure, only 42.06% recognized the disclosure. Although the disclosure was not correctly recognized by all participants, the manipulation was successful with a significant difference among groups, F(3, 429) = 74.66, p < .001.
Randomization Checks
The eight experimental groups did not differ with respect to sex, χ2(7) = 3.92, p = .789, age, F(7,425) = .77, p = .613, education level, F(7, 425) = .55, p = .794, frequency of Instagram use F(7, 425) = .53, p = .816, brand familiarity, χ2(7) = 4.15, p = .762, attitude towards Instagram F(7,425) = .29, p = .958, and product interest, F(7,425) = .53, p = .810. Effects of Disclosure or Cue and Influencer Type
To test H1 and H2, a factorial two-way ANOVA was conducted to assess the main effect of disclosure type and the interaction effect between disclosure cue and influencer type on ad recognition. The main effect for disclosure cue on ad recognition was significant, F(3, 425) = 38.66, p < .001 (see Table 3 for ad recognition means per disclosure cue condition). Post hoc pairwise comparisons using the Bonferroni correction demonstrated that ad
recognition was significantly higher with Instagram’s standardized disclosure (Mstandardized = 6.02, SDstandardized = 1.31) compared to both disclosure cue types (M#paidad = 5.06, SD#paidad = 1.47, p = .000; Mbrand tag = 5.12, SDbrand tag = 1.45, p = .000) and the absence of a disclosure (Mno disclosure = 3.98, SDbrand tag = 1.42, p = .000). Additionally, posts without any disclosure cue yielded a significantly lower ad recognition compared to both #paidad (p = .000) and brand tag (p = .000). Ad recognition was second highest for posts with a brand tag but not significantly different from the slightly lower ad recognition in the #paidad condition (p = 1.000). These results partly supported H1. Instagram’s standardized disclosure indeed lead to the highest ad recognition while the absence of a disclosure cue yielded the lowest ad
recognition. However, in contrast to the expectations, a brand tag resulted in a higher ad recognition than #paidad, although this difference was not statistically significant. Table 3
Ad Recognition Means per Disclosure Cue Type. Standardized
Disclosure #paidad Brand tag No disclosure or cue
Ad recognition 6.02 a 5.06b 5.12b 3.98c
Moreover, the ANOVA revealed that there was merely tentative support for a moderation of influencer type on the effect of disclosure cue type on ad recognition with a marginal significance level, F(3, 425) = 2.46, p = .062. Hence, H2 is not fully supported and moderation analyses were neglected in all subsequent analyses. Moreover, the results
indicated a main effect of influencer type on ad recognition, F(1, 425) = 12.22, p = .001, with significantly higher ad recognition for participants exposed to a macro-influencer (M = 5.29, SD = 1.59) compared to those exposed to a nano-influencer (M = 4.79, SD = 1.60). While influencer type did have a direct effect on ad recognition, it did not influence ad skepticism, influencer trustworthiness, brand attitudes or purchase intentions (see Appendix II, Table 3). Cognitive Consequences
Using PROCESS (Hayes, 2018), a logistic regression with disclosure cue type as a predictor, ad recognition as a mediator and brand recall as a dependent variable was conducted to test H3 (see Figure 4 for mediation model and Table 4 for corresponding results).
Figure 4. Tested Mediation Model: Effect of Disclosure Cue Type on Brand Recall via the Recognition of Influencer Marketing as Advertising.
The results in Table 4 show significant effects on ad recognition for all disclosure cue comparisons except for when #paidad was shown in the caption compared to a brand tag (ba = -.06, p = .756). This corresponds to the results found in the ANOVA previously conducted to test H1. Furthermore, the extent to which participants recognized the post as influencer marketing significantly increased brand recall (bb = .77, p = .000). Additionally, the analyses revealed that there was no direct effect of a disclosure cue on brand recall, except for when
#paidad was shown in the capture (compared to brand tag; bc = -.88, p = .005). Moreover, bootstrapping indicated significant indirect effects via the mediator ad recognition for all disclosure cues (indirect effectStandardized (No disclosure) 1.57, SE = .24, 95% BCI [1.18, 2.13]; indirect effectBrand tag (No disclosure) .88, SE = .18, 95% BCI [.57, 1.29]; indirect effect#paidad (No disclosure) .83, SE = .19, 95% BCI [.52, 1.26]; indirect effect#brand tag (standardized) -.69, SE = .17, 95% BCI [-1.08, -.40]; indirect effect#paidad (standardized) -.74, SE = .17, 95% BCI [-1.14, -.43]) except for when #paidad was present in the caption (compared to brand tag; indirect effect -.05, SE = .16, 95% BCI [-.35, .26]).
Table 4
Effect of Disclosure Cue Type on Brand Recall via Recognition of Advertising. Disclosure/Cue
(Reference) Indirect effect (SE) [95% BCI] a b c’
Standardized (no disclosure) 1.57 (.24) [1.18, 2.13] 2.04 (.19)*** .77 (.09)*** -.17 (.36) Brand tag (no disclosure) .88 (.18) [.57, 1.29] 1.14 (.19)*** . . . .16 (.32) #paidad (no disclosure) .83 (.19) [.52, 1.26] 1.08 (.19)*** . . . -.72 (.33) Brand tag (standardized) -.69 (.17) [-1.08, -.40] -.09 (.19)*** . . . .32 (.33) #paidad (standardized disclosure) -.74 (.18) [-1.14, -.43] -.96 (.19)*** . . . -.56 (.33) #paidad (brand tag) -.05 (.16) [-.35, .26] -.06 (.13) . . . -.88 (.31)***
Note. Unstandardized b-coefficients (with boot SE); BCI= bootstrap confidence interval using 5,000 bootstrap
samples; significant indirect effects are bold; ...= scores are the same as the scores above; standardized = Instagram’s standardized disclosure; N=433.
*p < .05, **p < .01, ***p < .001, †p < .10.
To summarize, the results support H3: compared to no disclosure, a disclosure cue always enhanced the recognition of advertising, which consequently increased brand recall. Comparing the disclosure cues to each other, brand tag and #paidad had a negative significant indirect effect compared to Instagram’s standardized disclosure, while #paidad (compared to brand tag) did not have a significant indirect effect. Hence, Instagram’s standardized
disclosure is significantly more effective in increasing brand recall via ad recognition than any other disclosure or cue while brand tag and #paidad do not differ among each other.
Evaluative Consequences
To test H4, three serial mediator models with disclosure cue type as the independent variable, ad recognition as first mediator, skepticism towards the ad as second mediator and either perceived influencer trustworthiness, brand attitudes or purchase intention as the dependent variable were tested using PROCESS (Hayes, 2018; see Figure 5 for model and Table 5 for corresponding results). The c-path in the model includes the direct effect of disclosure cue type on the respective dependent variable, independent of the effect of the mediators (c′), and the total effect of disclosure cue type on the dependent variable (c), which is the sum of the direct effect and the indirect effect via the mediators (Hayes, 2018). Results for path a1 are identical to the results for path a in Table 4 and are, thus, not further discussed.
Figure 5. Tested Serial Mediation Model: Effect of Disclosure Cue Type on Perceived Influencer Trustworthiness, Brand Attitudes and Purchase Intention via Recognition of Influencer Marketing as Advertising and Skepticism towards Advertising.
In line with H4a, the results demonstrate that the extent to which participants recognized a post as advertising significantly increased skepticism towards the commercial content (ba3 = .09, p = .028). In addition, ad skepticism had a significant negative effect on perceived influencer trustworthiness (bb2 = -.33, p = .000), brand attitudes (bb2 = -.30, p = .000) and purchase intention (bb2 = -.48, p = .000), in support of H4b, H4c and H4d.
Moreover, the findings indicate that both the total (c) and the direct effect (c’) were not significant for any of the disclosure cues, respectively any of the dependent variables. Hence, there was no direct effect of any disclosure cue on perceived influencer
trustworthiness, brand attitudes or purchase intention. Additionally, bootstrapping revealed significant indirect effects via the two mediators for all disclosure cues compared to the absence of a disclosure cue on influencer trustworthiness (indirect effectStandardized (No disclosure) -.06, SE = .03, 95% BCI [-.12, -.01]; indirect effectBrand tag (No disclosure) -.03, SE = .02, 95% BCI [-07, -.004]; indirect effect#paidad (No disclosure) -.03, SE = .02, 95% BCI [-.07, -.004]), brand attitude (indirect effectStandardized (No disclosure) -.06, SE = .03, 95% BCI [-.11, -.01]; indirect effectBrand tag (No disclosure) -.03, SE = .02, 95% BCI [-.06, -.003]; indirect effect#paidad (No disclosure) -.03, SE = .01, 95% BCI [-.06, -.003]) and purchase intention (indirect effectStandardized (No disclosure) -.09, SE = .04, 95% BCI [-.17, -.01]; indirect effectBrand tag (No disclosure) -.05, SE = .02, 95% BCI [-.10, -.01]; indirect effect#paidad (No disclosure) -.05, SE = .02, 95% BCI [-.09, -.01]). Thus, the presence of a disclosure cue always increased ad recognition compared to its absence. Consequently, participants were more skeptical towards the ad, which resulted in lower perceptions of the influencer’s trustworthiness, less favorable brand attitudes and a lower likelihood to purchase the advertised product. Analyses comparing the different types of disclosure cues showed a significant indirect effect of brand tag and #paidad compared to Instagram’s standardized disclosure on influencer trustworthiness (indirect effectBrand tag (standardized) .03, SE = .01, 95% BCI [.004, .06]; indirect effect#paidad (standardized) .03, SE = .01, 95% BCI [.003, .06]), brand attitudes (indirect effectBrand tag (standardized) .02, SE = .01, 95% BCI [.003, .05]; indirect effect#paidad (standardized) .03, SE = .01, 95% BCI [.003, .06]) and purchase intention (indirect effectBrand tag (standardized) .04, SE = .02, 95% BCI [.01, .08]; indirect
effect#paidad (standardized) .04, SE = .02, 95% BCI [.005, .09]). However, there was no serial mediation for #paidad compared to brand tag on influencer trustworthiness (indirect
effect#paidad (brand tag) .002, SE = .01, 95% BCI [-.01, .02]), brand attitudes (indirect effect#paidad (brand tag) .002, SE = .01, 95% BCI [-.01, .02]) and purchase (intention indirect effect#paidad (brand tag) .003, SE = .01, 95% BCI [-.02, .02]).
Table 5
Effect of Disclosure Cue Type on Influencer Trustworthiness, Brand Attitudes and Purchase Intention via Recognition of Advertising and Skepticism Towards the Ad.
Dependent Variable: Influencer Trustworthiness Disclosure cue
(Reference) (SE) [95% BCI] Indirect effect a1 a2 a3 b1 b2 (total) c (direct) c’ Standardized (no disclosure) [-.12, -.01] -.06 (.03) (.19)2.04 *** .32 (.19)† (.04).09 * (.03) -.03 (.04)-.33 *** (.14) .07 (.14) .09 Brand tag (no disclosure) [-.07, -.004] -.03 (.02) (.19)1.14 *** (.17) .09 . . . . . . . . . (.13) -.10 (.13) -.06 #paidad (no disclosure) [-.07, -.004] -.03 (.02) (.19)1.08 *** (.17).35 * . . . . . . . . . (.14) .01 (.13) -.04 Brand tag (standardized) [.004, .06] .03 (.01) (.19)-.90 *** (.17) -.23 . . . . . . . . . (.14) -.17 (.13) -.15 #paidad (standardized) [.003, .06] .03 (.02) (.19)-.96 *** (.17) .03 . . . . . . . . . (.14) -.06 (.13) -.13 #paidad (brand tag) [-.01, .02] .002 (.01) (.19) -.06 (.17) .26 . . . . . . . . . (.14) .11 (.12) .02 Dependent Variable: Brand Attitudes
Standardized (no disclosure) [-.11, -.01] -.06 (.03) (.19)2.04 *** .32 (.19)† (.04).09 * (.03) -.03 (.03)-.30 *** (.12) -.14 (.12) -.13 Brand tag (no disclosure) [-.06, -.003] -.03 (.02) (.19)1.14 *** (.17) .09 . . . . . . . . . (.12) -.13 (.11) -.10 #paidad (no disclosure) [-.06, -.003] -.03 (.01) (.19)1.08 *** (.17).35 * . . . . . . . . . (.12) .02 (.11) -.07 Brand tag (standardized) [.003, .05] .02 (.01) (.19)-.90 *** (.17) -.23 . . . . . . . . . (.12) .01 (.11) .03 #paidad (standardized) [.003, .06] .03 (.01) (.19)-.96 *** (.17) .03 . . . . . . . . . (.12) .12 (.11) .06 #paidad (brand tag) [-.01, .02] .002 (.01) (.19) -.06 (.17) .26 . . . . . . . . . (.12) .11 (.11) .03 Dependent Variable: Purchase Intention
Standardized (no disclosure) [-.17, -.01] -.09 (.04) (.19)2.04 *** .32 (.19)† (.04).09 * (.03) -.03 (.06)-.48 *** -.003 (.21) (.21) -.03 Brand tag (no disclosure) [-.10, -.01] -.05 (.02) (.19)1.14 *** (.17) .09 . . . . . . . . . (.20) -.13 (.20) -.10 #paidad (no disclosure) [-.09, -.01] -.05 (.02) (.19)1.08 *** .35 (.17)* . . . . . . . . . -.02 (.21) (.20) -.11 Brand tag (standardized) [.01, .08] .04 (.02) (.19)-.90 *** (.17) -.23 . . . . . . . . . (.21) -.12 (.20) -.07 #paidad (standardized) [.005, .09] .04 (.02) (.19)-.96 *** (.17) .03 . . . . . . . . . (.21) .01 (.20) -.09 #paidad (brand tag) [-.02, .02] .003 (.01) (.19) -.06 (.17) .26 . . . . . . . . . (.21) .11 (.19) -.02
Note. Unstandardized b-coefficients (with boot SE); BCI= bootstrap confidence interval using 5,000 bootstrap
samples; significant indirect effects are bold; …= scores are the same as the scores above; standardized = Instagram’s standardized disclosure; N=433.
Summarizing, the results support H3a, H3b, H3c, and H3d: compared to no
sponsorship disclosure or cue, a disclosure cue can increase the recognition of advertising, which evokes skepticism towards the commercial post, and eventually results in a negative indirect effect on influencer trustworthiness, brand attitudes and purchase intention. Relating the disclosure cues to each other, only #paidad did not have a significant indirect effect compared to a brand tag. Instagram’s standardized disclosure, however, was significantly more effective than any other disclosure cue.
General Discussion and Conclusion
The present thesis investigated to what extent disclosure cues in an Instagram post and different types of influencers affect users’ ad recognition, their brand recall and
attitudinal reactions by conducting two experiments. More specifically, an eye-tracking study explored whether Instagram users recognize influencer marketing on Instagram and which disclosures or cues they use to do so. A subsequent online experiment examined to what extent the disclosure cues identified as helpful in the first study, affect participants’ recognition of advertising and further cognitive and evaluative responses towards the persuasive message, the influencer and the advertised brand. Additionally, it was assessed whether influencer type functions as a moderator in the effect of disclosure cues on ad recognition.
First, the eye-tracking findings grant practical insights by showing that individuals are able to recognize particularly those posts who feature Instagram’s standardized disclosure or were posted by internationally known ‘Instagrammers’. However, the results also suggest that one in three commercial posts remained unrecognized, a finding that differs from previous studies on social media advertising that detected ad recognition rates of more than 80% (Johnson et al., 2019, Jung & Heo, 2019, Kim & Song, 2018). This discrepancy may be arise due to the present study’s more realistic experimental design, in which participants were exposed to a video showing 50 posts, instead of displaying only one post, which is the case in
most other studies. It may also be attributable to the present study’s consideration of vague cues, such as a plain brand tag, that were neglected in prior research. These cues are less explicit than sponsorship disclosures and, may, thus results in lower ad recognition. Including such cues in studies on sponsorship disclosures on social media is crucial since it is common practice among influencers to ‘disclose’ commercial posts with such ambiguous cues
although they are not considered as sufficient disclosures by regulatory parties. Hence, this study adapts to today’s reality by including diverse disclosure cues and reveals that users lack sufficient persuasion knowledge to recognize influencer marketing without difficulty.
Second, participants’ eye movements reveal that Instagram’s standardized disclosure as well as brand tags in the picture catch users’ attention longer than all other tested
disclosure cues and help almost half of the participants to identify ads. Additionally, the online experiment shows that Instagram’s standardized disclosure increased ad recognition more effectively than the other tested disclosure cues. These results support the notion that, when a disclosure conveys both the paid relationship and the sponsor in a clear and direct manner, consumers are provided with information that allows them to accurately interpret the commercial nature of the post (Evans et al., 2017; Hyman et al., 2017; Wojdynski et al., 2017). Furthermore, a brand tag ranks second in increasing ad recognition, even slightly but not significantly more effective than #paidad. Thus, a cue that is not acknowledged by FTC as a sufficient disclosure (FTC, 2017a, 2017b), appears to be a highly noticed and helpful element for Instagram users to identify ads. These finding suggest that not only disclosures but also cues are highly effective in activating persuasion knowledge and, consequently, affecting consumers’ cognitive and evaluative responses. This presents consumers with a problem: Such cues are not necessarily used only when posts are commercial. Hence, people might misidentify non-commercial posts that tag or mention brands as advertising. In line with this notion, the eye-tracking results indicated that almost one in five non-commercial posts is erroneously assumed to be advertising. However, this issue does not only affect
consumers. Brands that are tagged in posts without paying the ‘Instagrammer’ to do so might dissent this ostensible affiliation. For instance, brands might rather not be associated with particular ‘Instagrammers’, who do not fit their brand.
Furthermore, this study’s results demonstrated that influencer marketing with brand tags can harm a brand since it lowers consumers’ brand attitudes and purchase intention. Hence, brands may disapprove of ‘Instagrammers’ tagging the brand when there is no paid partnership because they apprehend that users may perceive the brand tag as a persuasive attempt and, thus, react negatively. Similarly, ‘Instagrammers’ should consider that tagging a brand – even without having a commercial relationship with that brand – may result in lower perceptions of the influencers’ trustworthiness.
Third, influencer type directly influences ad recognition, with a significantly higher recognition for posts by macro-influencers compared to nano-influencers. While the prominence of an ‘Instagrammer’ functions as a cue that stimulates persuasion knowledge, there is merely a tentative support for a moderating effect of influencer type on the impact of disclosure or cue on ad recognition. Influencer type does not directly affect ad skepticism, influencer trustworthiness, brand attitudes and purchase intention, albeit these responses are slightly but not significantly more positive when a post is sent by a nano- compared to a macro-influencer.
Fourth, the tested disclosure cues were found to elicit a cognitive process, in which brand recall is improved via ad recognition. Any type of disclosure cue resulted in a higher brand recall compared to the absence of it, with Instagram’s standardized disclosure being the most effective disclosure followed by a brand tag. This suggests that a disclosure emphasizes the advertised brand and is, thus, more likely to be noticed by consumers since they use their limited capacity to process it (Boerman et al., 2014; Lang, 2000). Among the disclosures, #paidad had the lowest effect on brand recall, suggesting that it is easier for users to recall the brand when disclosure cues mention it, as with the standardized disclosure and brand tag.
Fifth, the disclosure cues were also found to evoke an evaluative process. Consistent with the PKM, the results demonstrate that the presence of a disclosure and the resulting increase of ad recognition and skepticism elicit a more critical reaction towards the post, the influencer and the advertised brand. While the indirect negative effects of disclosures on influencer trustworthiness and brand attitudes replicate previous findings by De Veirman and Hudders (2019), this study is the first to demonstrate disclosure cues’ negative influence on purchase intention via ad recognition and skepticism in the context of Instagram. Hence, disclosure cues do not only affect consumers’ evaluative reactions but also their behavioral intentions. Moreover, the present thesis adds to current knowledge by revealing that this ‘change of meaning’ does not only occur when a disclosure is displayed but also when an Instagram post features a brand tag. Henceforth, merely using a brand tag is similarly harmful for the influencer’s trustworthiness and the brand as explicitly disclosing influencer
marketing and, thus, does not yield potentially supposed advantages for ‘Instagrammers’ or brands.
These findings have implications for various stakeholders. Regulatory parties, such as the FTC previously denied Instagram’s standardized disclosure’s sufficiency (FTC, 2017b) and, thus, understated its utility. In contrast, the present studies suggest that Instagram’s standardized disclosure is considerably more attention-grabbing and effective in activating persuasion knowledge than other disclosures, including the FTC-respected #paidad.
Additionally, even brand tags, which are not accepted as a disclosure by regulatory parties, increase ad recognition slightly, although not significantly, more than the seemingly clearer #paidad. Therefore, regulatory parties may need to rethink their classification of adequate sponsorship disclosures.
The fact that consumers have developed sufficient PK to not only understand unambiguous disclosures but to also interpret cues as disclosures initially appears to be beneficial. However, unquestioningly assuming that cues such as a brand tag always reveal a
paid relationship between ‘Instagrammer’ and brand may frequently lead to an erroneous classification of non-commercial posts as advertising. This demonstrates the blurred lines of advertising on Instagram and reveals grey areas in which posts are perceived as advertising, even though they are not and vice versa. Therefore, the results of the present study call for clearer recommendations on behalf of the FTC on what influencers should use as disclosures, but also which elements they should relinquish.
For advertisers, the results demonstrate that their choice to partner up with macro- or nano-influencers does not have a direct effect on consumers’ brand attitudes and purchase intentions. However, posts sent by macro- compared to nano-influencers do increase ad recognition, which in turn affects evaluative brand responses negatively via ad skepticism. Thus, advertisers should carefully consider whether positive effects of working with macro-influencers, such as their reach, outweigh the negative consequences that emerge when consumers realize that a post is advertising. In contrast, when brands partner up with nano-influencers, users are less likely to recognize the ad and, thus, less likely to provoke negative reactions.
Moreover, ‘Instagrammers’ should carefully decide if or how extensively they engage in influencer marketing since sponsorship disclosures were found to have a negative impact on users’ perceptions of their trustworthiness via ad recognition and ad skepticism. This affects macro-influencers more strongly than nano-influencers cause consumers are more likely to recognize advertising when it is posted by popular ‘Instagrammers’.
To conclude, some limitations of the current thesis, that may provide further guidance for future research, are discussed. First, the data conducted in the eye-tracking experiment did not allow for an analysis of the effect of disclosure cue type on ad recognition, potentially mediated by visual attention, due to its design. More specifically, each post contained a combination of a disclosure and a cue and there were few posts featuring the same
