Making and breaking relationships on social media : the impacts of brand and influencer betrayals

This study considers how the relationships between social media influencers, brands and individuals are inter-twined on social media and analyses the spill-over effects of feelings of betrayal. An experimental design with two transgression scenarios (influencer vs. brand) was created, and 250 individuals were recruited to participate in the study. The results show that a perceived betrayal by a brand can negatively affect the perceived coolness of the social media influencer that has endorsed the brand, as well as the parasocial relationships that followers have with the influencer. Accordingly, a perceived betrayal by a social media influencer can negatively affect attitudes, trust and purchase intentions toward a brand that the influencer has endorsed. The current research helps in understanding brand and influencer transgressions and highlights the fact that both influencers and brands should have a sense of collaboration responsibility. It also introduces the concept of influencer coolness, understood here as a desirable success factor for social media influencers, which partly explains their desirability and influence, and a feature that can be endangered through both influencer and brand betrayals.


Introduction
The benefits of social media are numerous, including the fact that social media platforms enable new possibilities to interact and maintain relationships with friends and loved ones (e.g., boyd and Ellison, 2007;Cheung et al., 2011;Dhir et al., 2018;Gennaro and Dutton, 2007). In addition, social media also allows connecting with new kinds of 'digital friends': These new types of relationship partners can include social media influencersbloggers, YouTubers, Instagram and TikTok celebritieswho are often considered close friends or even family members by their followers (Berryman and Kavka, 2017;Reinikainen et al., 2020), and also brands (Fournier, 1998), which can evoke strong self-brand connections (Tan et al., 2019) or even feelings of love (Batra et al., 2012) in devoted individuals. This effect may be especially high in young people, who are often 'fascinated with popular brands' (Dhir et al., 2016: 427).
These intimate, online relationships have become intertwined as brands seek endorsements from 'cool' influencers (Ember, 2015) to gain the trust and attention of the influencers' followers. For instance, within the beauty industry, influencers, followers and brands have constituted active online communities (Lawson, 2021). Such collaborations contribute to many positive outcomes for brands through, for example, heightened purchase intentions (Lee and Watkins, 2016) and, for influencers, a chance to enhance connectedness with followers (Reinikainen et al., 2020). For the public at large, the contribution of influencer-brand collaborations remains controversial, but for devoted followers, the collaborations offer a mechanism via which to ensure that the influencers receive incentives in exchange for their work and, thus, followers can continue receiving their favourite influencer content (Coco and Eckert, 2020).
However, mismanaging the relationships between influencers, followers and brands can also lead to negative emotions. Feelings of intimacy and closeness can turn into feelings of betrayal when moral obligations are broken or integrity is lost (Tan et al., 2021). Moreover, research on online communities suggests that negative experiences and emotions may spill over from the original target to another target (Bowden et al., 2017) and discourage participation in these communities via online regret . Many influencers have, therefore, become vigilant in terms of entering collaborations with brands and deliberately select partners that are consistent with their personal brands (Watson, 2020) to maintain their trustworthiness and coolness in the eyes of their followers (Ember, 2015). Similarly, brands show concern over risking their reputation in entering endorsement deals with influencers who may exhibit questionable behaviours on social media (Winchel, 2018). For instance, American YouTuber Logan Paul caused controversy by filming a dead body in a Japanese suicide forest (Farokhmanesh, 2018), leading to furiously negative feedback and YouTube removing Logan Paul as a preferred advertisement partner. This suggests that, although the collaborations between influencers and brands are often profitable for both parties, potential transgressions by either party that lead to feelings of betrayal may have negative effects on the other party while also causing feelings of distress and uneasiness for followers.
Most of the research on social media influencers and brands has concentrated on the positive outcomes of influencer endorsements on individual attitudes and engagement in sponsoring brands (e.g., De Veirman et al., 2017;Dhanesh and Duthler, 2019;Hughes et al., 2019). However, prior academic research has largely ignored the spill-over effects of influencer transgressions on endorsed brands, although the impact of endorser scandals are a common topic in the context of mainstream celebrities (Bartz et al., 2013;Carrillat et al., 2013;Carrillat, et al., 2014;Louie et al., 2001;Till and Shimp, 1998). In the same vein, there is limited research on the spill-over effect of brand transgressions on social media influencers; most brand transgression research focuses on the impacts on firm and individual perspectives (Khamitov et al., 2019).
In the present study, we aim to fill this research gap by asking how a perceived betrayal by a brand that is endorsed by a social media influencer is connected with the followers' relationship with that influencer and, vice versa, how a perceived betrayal by an influencer endorsing a brand affects individuals' brand attitudes, trust and purchase intentions. We also introduce the concepts of influencer coolness and influencer betrayal. Influencer coolness is derived from the concept of brand coolness (Warren et al., 2019), understood here as a desirable success factor not just for company brands but also for human brands, such as social media influencers. Influencer betrayal is derived from the concept of brand betrayal (Reimann et al., 2018), understood here as a feeling of betrayal that may be reflected in the relationship between a follower and social media influencer.
Our findings have significant practical implications for the understanding of collaboration responsibility from the perspective of both brands and influencers. This responsibility implies a shift away from the typical self-care approach, in which the parties are mostly worried about their reputation, and toward an understanding that influencers and brands have responsibilities regarding one another's reputation.
In the next section, we review the literature on social media influencers, parasocial relationships, influencer coolness and betrayal to develop our conceptual framework and hypotheses. This is followed by a presentation of the results obtained from the experiment testing of our hypotheses. Finally, we discuss the theoretical contributions, practical implications, limitations and suggestions for future research.

Social media influencers
Social media influencers, especially their ability to address and impact their followers, have become a growing interest topic in both practice and academic research. Social media influencers have been defined as 'third-party endorsers who shape audience attitudes through blogs, tweets, and the use of other social media' (Freberg et al., 2011: 90). Other essential features that define influencers include direct and active interaction with followers; professional and effective content production and distribution skills (Enke and Borchers, 2019); the ability to monetise their following through, for example, sponsored material within their content (Abidin, 2016) and personal branding (Dhanesh and Duthler, 2019).
Although many of the most popular Youtubers and Instagrammers are as well-known as mainstream celebrities, such as pop singers or movie stars, it is important to distinguish between social media influencers and so-called traditional celebrities. Influencers have often found their way to fame on their own through social media and bypassed the gatekeeper role of the mass media (Hou, 2019), whereas so-called mainstream celebrities often depend heavily on the attention of traditional mass media (Enke and Borchers, 2019). Another difference between mainstream celebrities and social media influencers can be found in the different ways they relate to their fans and followers (Enke and Borchers, 2019): while mainstream celebrities often strive to keep a distance from audience members (Jerslev, 2016) and value their privacy, the appeal of influencers lies in their openness in terms of sharing their most intimate feelings and life events with their followers (Marôpo et al., 2020).
Because of these intimate and influential relationships that influencers have with their followers, co-operation with influencers has been embraced by brands, which are able to profit from the collaborations, for example, through heightened purchase intentions (Lee and Watkins, 2016;Sokolova and Kefi, 2020), brand attitudes (De Veirman et al., 2017;Munnukka et al., 2019), brand trust (Reinikainen et al., 2020) and eWOM intentions (Hwang and Zhang, 2018). The effectiveness of influencer endorsements lies partially in the friend-like parasocial relationships that followers have with their favourite influencers (Lee and Watkins, 2016;Reinikainen et al., 2020), but source characteristics, such as the perceived credibility (Munnukka et al., 2019), authenticity (Pöyry et al., 2019) and attractiveness (Wiedmann and von Mettenheim, 2020) of the influencer, have also been deemed important.

Parasocial relationships
A parasocial relationship refers to the illusion of a face-to-face relationship with a person who is encountered through media (Horton and Wohl, 1956). Originally a concept coined to describe the relationships that audience members develop with performers on television, radio and film, it has since been applied in studying influencers who are encountered through various social media channels (Colliander and Dahlén, 2011;Lueck, 2015;Lee and Watkins, 2016;Yuan and Lou, 2020). Parasocial relationships resemble social relationships in many ways; however, they are often not balanced, because, although the audience members know a great deal about the performer, the reverse is not true (Munnukka et al., 2019). The experience of parasocial relationships may entail the feeling of knowing the performer well (Brown, 2015), identification with and interest in the performer (Auter and Palmgreen, 2000), the feeling of being among friends while consuming media content and wanting to meet the performer in real life (Rubin et al., 1985).
Parasocial experiences have been shown to affect the attitudes and behaviours of audience members in many ways (Tian and Hoffner, 2010). In the context of social media influencers, parasocial relationships have been found to advance the effectiveness of influencer endorsements through, for example, mediating the relationship between audience participation and influencer credibility (Munnukka et al., 2019) and the relationship between source credibility and product interest (Yuan and Lou, 2020). Parasocial relationships also moderate the effect of persuasion knowledge (Hwang and Zhang, 2018), suggesting the effectiveness of endorsements made by 'digital friends'.

Influencer coolness
Individuals desire 'cool' things and are willing to spend money on 'cool' brands (Warren et al., 2019). Coolness has been defined as 'a subjective and dynamic, socially constructed positive trait attributed to cultural objects inferred to be appropriately autonomous' (Warren and Campbell, 2014: 544). According to Warren et al. (2019), cool brands are extraordinary, aesthetically appealing, energetic, original, authentic, rebellious, high status, subcultural, iconic and popular, and increasing any one of these characteristics tends to make a brand seem cooler.
In previous studies, coolness has been connected with brands such as Harley-Davidson and Apple (Warren and Campbell, 2014) but also with celebrities such as Jay-Z and Beyoncé (Warren et al., 2019). In the current study, we examine whether the concept of brand coolness could also be applied in the context of social media influencers, often defined as 'human brands' (Delisle and Parmentier, 2016;Dhanesh and Duthler, 2019). Coolness has been seen as a desirable feature for brands, so the present study perceives coolness as a feature that is also desirable for social media influencers in that it attracts not only followers but also brands that wish to be associated with cool influencers.
Following the notion of Warren et al. (2019), we define influencer coolness as the perceived level of the influencer demonstrating both desirability and positive autonomy by being original, energetic, appealing, extraordinary, high status, iconic and popular. In this sense, coolness covers a holistic view of socially constructed positive traits attributed to a social media influencer, moving beyond the investigation into other source characteristics, such us the perceived attractiveness (Wiedmann and von Mettenheim, 2020), credibility (Reinikainen et al., 2020), expertise (Trivedi and Sama, 2020) and popularity (De Veirman et al., 2017) of the influencer.
The current study focuses on influencer coolness for two reasons. First, the four key features of coolnesssubjectivity, positivity, autonomy and dynamicityas defined by Warren and Campbell (2014), are also descriptive of social media influencers. Influencers are only as cool as their followers perceive them to be, and transgressions performed by influencers are difficult for their followers to come to terms with (Cocker et al., 2021). Social media influencers are also often perceived positively by their followers (Reinikainen et al., 2020), and they have autonomy, but they are also dynamic and have been discovered to follow 'an evolutionary process' or lifecycle, in which their relationship with their followers changes over time (Smith, 2010). Second, coolness is associated with behavioural outcomes, such as intentions to talk about a brand and willingness to pay for the brand (Warren et al., 2019). Similar intentionswillingness to share and engage with an influencer's post and pay for products that are endorsed by the influencerare also desirable for social media influencers. Coolness is, therefore, understood here as a success factor for social media influencers that, in part, explains their desirability and also their effectiveness as endorsers.

Brand and influencer transgressions
The effects of celebrity endorser scandals on brands have been studied quite frequently within the fields of marketing and management studies. These studies show that transgressions on the part of celebrity endorsers, such as the use of illegal substances or other behaviours that lead to negative headlines, negatively impact attitudes toward the endorsed brands (Till and Shimp, 1998) and the financial performance of these brands (Bartz et al., 2013;Hock and Raithel, 2019). Moreover, Carrillat et al. (2014) showed that a celebrity endorser's wrongdoing may impact not only the attitudes toward the endorsed brand but also other competing brands. The mechanism also seems to work conversely: Kelly et al. (2016) showed that negative perceptions about a sports team can lead to negative perceptions about the sponsoring brand. Thomas and Fowler (2016) also pointed out that the transgressions made by brands can have an effect on how individuals evaluate celebrities who have endorsed these brands.
Corresponding evidence from the context of social media influencers is still scarce, although controversies involving both influencers and brands are common and recognised in the context of influencer endorsements (Cocker et al., 2021;Lawson, 2021). Because the relationships that fans and followers have with social media influencers differ somewhat from the relationships they have with mainstream celebrities, the dynamics of brand and influencer transgressions and their effects on influencers and brands, respectively, constitute a gap in the research literature. Previously, Colliander and Erlandsson (2015) discovered that exposing a hidden sponsorship between a blogger and brand negatively affects followers' attitudes toward the blog and the credibility of the blogger, with the parasocial relationship with the blogger mediating the effect. However, they were not able to detect a significant negative impact on brand attitudes and purchase intentions. Because this is somewhat contradictory to the results of studies on celebrity transgressions, further research is called for, as well as the introduction of additional constructs that may be essential to the desirability of influencers, but that Colliander and Erlandsson (2015) may not have considered in their study.

Brand and influencer betrayal
The concept of brand betrayal has inspired academic research over the past years because there has been an increasing tendency toward humanising brands and seeing them as relationship partners that are close to individuals (Fournier, 1998;Tan, 2018). Individuals form expectations for their favourite brands, and if these expectations fail and brands are found guilty of transgressions, individuals can experience feelings of betrayal (MacInnis and Folkes, 2017;Reimann et al., 2018), much like in other relationships.
Brand betrayal has been defined as an unpleasant emotion evoked by a moral violation on the part of a brand that the individual has a strong self-brand connection with, fracturing the relationship that the individual has with the brand (Reimann et al., 2018;Tan et al., 2021). Brand betrayal has been deemed one of the most intrinsic factors leading individuals to behave negatively toward the brand (MacInnis and Folkes, 2017). The experience of brand betrayal is composed of assessments of the feeling of being taken advantage, misled and exploited by a brand (Tan, 2018).
Due to the close ties that followers have with influencers, it seems likely that feelings of betrayal could also follow transgressions on the part of influencers. Following the work of Tan et al. (2021), in the current study, influencer betrayal is defined as an unpleasant feeling caused in a follower by a moral violation on the part of a social media influencer, and it could have ramifications for the relationship that followers have with the influencer. The feeling of being misled and exploited is considered essential to the experience.

Conceptual framework and hypotheses
Based on the reviewed literature, we propose that a negative spillover effect exists 1) on the endorsed brand when individuals feel betrayed by an influencer they follow and 2) on the endorsing influencer when individuals feel betrayed by a brand they favour. The following conceptual frameworks and related hypotheses, which present the spillover effects from influencer transgression on the endorsed brandand vice versaare illustrated in Figure 1.
Previous studies show that parasocial relationships can influence not only the opinions, interests and attention of people on social media but also the construction of relationships (Yuksel and Labreque, 2016). Therefore, the parasocial relationship is considered a key construct in the current model. Because parasocial relationships have been found to resemble social relationships (Rubin and McHugh, 1987), we suggest that parasocial relationships can be affected by negative, relational experiences, such as influencer betrayal: H1. Following an influencer transgression, influencer betrayal negatively affects the parasocial relationship with the influencer.
Another key construct in our model is influencer coolness, defined in this study as an important success factor for influencers, explicating their desirability and ability to convince their followers. Because brand coolness has been connected with the emotional connection of brand love (Warren et al., 2019), we suggest that influencer coolness is positively related with an emotional connection with an influencer: H2. Parasocial relationship is positively related with influencer coolness.
Here, we treat influencer coolness as a source characteristic construct. Colliander and Erlandsson (2015) found that a transgression made by a blogger had a negative effect on the perceived credibility of the blogger and that a parasocial relationship with the blogger mediated this relationship. As we view both influencer coolness and credibility as source characteristic constructs, we suggest the following hypothesis, which follows the same line of dynamics as the study of Colliander and Erlandsson (2015): H3. Following an influencer transgression, a parasocial relationship mediates the relationship between influencer betrayal and influencer coolness, where influencer betrayal has negative and indirect effects on influencer coolness.
Source characteristics, such as perceived influencer credibility, have been shown to have a positive association with brand attitudes in the context of influencer marketing (Munnukka et al., 2019). We therefore suggest that influencer coolness is also positively associated with brand attitude: H4. Influencer coolness is positively related to endorsed brand attitude.
Brand attitude is often considered one of the most critical assets for brands because attitudes indicate behaviour (Kumra, 2007: 172). Another essential element for brands is brand trust, which reduces the uncertainty that individuals may feel toward a brand (Chaudhuri and Holbrook, 2001) and helps people make decisions regarding brands . Brand attitude has been found to be positively related to brand trust (Kim et al., 2019), and both brand attitude and brand trust have been found to be positively associated with purchase intention in influencer marketing (Munnukka et al., 2019;Reinikainen et al., 2020). Thus, we suggest the following two hypotheses: H5. Brand attitude is positively related to brand trust.

H6.
Brand trust is positively related to purchase intention.
When a transgression occurs that leads to feelings of betrayal, a negative brand attitude is expected to result, producing a lower level of brand trust (Folse et al., 2013). For this reason, we also suggest negative indirect effects on the part of influencer betrayal on brand attitude, brand trust and purchase intention: H7. Following an influencer transgression, influencer betrayal has negative and indirect effects on a) brand attitude, b) brand trust and c) purchase intention toward the endorsed brand.
Because brand betrayals have been found to fracture relationships between consumers and brands (Reimann et al., 2018), even leading to negative behaviours toward brands when there is a lack of brand recovery effort (MacInnis and Folkes, 2017;Tan et al., 2021), consumers are expected to have lower levels of positive attitudes and trust toward a betraying brand. For this reason, we suggest the following two hypotheses: H8. Following a brand transgression, brand betrayal negatively affects brand attitude.
H9. Following a brand transgression, brand attitude mediates the relationship between brand attitude and brand trust, where brand betrayal has negative and indirect effects on brand trust.
Because a trusted, positive relationship with a social media influencer is connected with a positive trusting relationship with a brand the influencer has endorsed (Reinikainen et al., 2020), we suggest that this effect may also function in the reverse direction: feelings of trust toward a brand should be connected with perceptions about the endorsing influencer and the relationship with said influencer. Therefore, the following two hypotheses are suggested: H10. Brand trust is positively related to influencer coolness.

H11.
Influencer coolness is positively related to a parasocial relationship with the influencer.
Based on the previous relationships, we also suggest our final hypothesis about the negative indirect effects of brand betrayal on purchase intention, influencer coolness and parasocial relationships. Our argument here is that, when consumers feel betrayed by a brand they have trusted, they are less likely to purchase or revisit the brand without a recovery effort (MacInnis and Folkes, 2017;Tan et al., 2021). In the same vein, these betrayed consumers should perceive a lower level of coolness with regard to a social media influencer who has endorsed the betraying brand. Consequently, they should be less likely to maintain a parasocial relationship with said influencer. Thus, we propose the following hypothesis: H12. Following a brand transgression, brand betrayal has negative and indirect effects on a) purchase intention, b) influencer coolness and c) the likelihood of a parasocial relationship with the influencer.

Method
An online survey was created using Qualtrics. Two hundred and fifty adults from across the US (106 men; age 19-74, M age = 34) were recruited from Amazon Mechanical Turk to participate in the study in exchange for a small reward. The participants were informed that the study was interested in the psychological factors that affect individuals' attitudes. They were first asked to list a social media influencer who they preferred and frequently followed on social media. Next, the respondents were asked to list a favourable brand that they had used. The respondents were then asked about the likelihood of an influencer endorsement through the following question: 'In your opinion, how likely is it that influencer X would endorse and recommend brand X on social media?' (1 = not at all likely to 7 = very likely) (M = 4.98 vs. midpoint 4; t(249) = 8.38, p < .001). After that, the respondents were randomly assigned to one of the two following transgression scenarios: Influencer transgression. Imagine that it is revealed to you that the listed influencer has bought followers for himself/herself on Instagram, YouTube and other social media channels. The listed influencer intentionally misleads you by positioning himself/herself as an authentic and popular influencer. With the help of these paid, fake followers, the listed influencer has pretended to be more popular than he/she is and tried to attract interest and sponsorships from brands to maximise personal income. For your information, the listed brand has been endorsed by the listed influencer in several social media posts. Brand transgression. Imagine that it is revealed to you that the listed brand intentionally misleads you by positioning itself as more ethical than its competitors. The truth is that the listed brand has been avoiding its responsibility to society by neglecting its code of conduct and ethical code, specifically participating in unethical practices, such as neglecting to ensure the safety of its employees and customers. Further, the brand aims to maximise its profit without treating its employees fairly. Also, the brand failed to inform you of a data breach, choosing instead to reward the hackers who had access to the customers' information to cover up its incompetence, giving away information such as your demographics and behavioural data.
For your information, the listed brand is endorsed by the listed influencer in several social media posts. Figures 1A and 1B, we propose that individuals travel different kinds of paths during the psychological process, starting from either the influencer transgression or brand transgression scenario. Thus, after reading the scenario, the respondents were asked to answer a manipulation check question: "Have you personally experienced an influencer/a brand that intentionally misled, took advantage of or exploited you as somehow presented in the given scenario?" (0 = No; 1 = Yes). Next, the respondents in the influencer transgression condition were asked to answer items about influencer betrayal (Tan et al., 2021), parasocial relationships (Munnukka et al., 2019), influencer coolness (Warren et al., 2019), brand attitude (Priester and Petty, 2003), brand trust (Chaudhuri and Holbrook, 2001) and purchase intention (Söderlund and Ö hman, 2003). The respondents in the brand transgression condition were asked to indicate their assessments of brand betrayal, brand attitude, brand trust, purchase intention, influencer coolness and a parasocial relationship. Apart from the one word interchanged between the items on influencer betrayal and brand betrayal, all other items were similar across the two conditions. All items were rated on a seven-point Likert scale from 1 to 7.

As pictured in
To check for manipulation, the respondents in the influencer (vs. respondents in the brand transgression condition indicated that they had encountered similar betrayals (85 per cent; χ 2 (63.48), p < 0.001). We decided to include all responses for the subsequent analysis because there were no statistically significant differences regarding participation time (p > 0.10) and the investigated constructs (p > 0.05).

Results
As presented in Table 1, the measurement models showed sufficient reliability and validity, as recommended by Hair et al. (2010) (Table 2). Table 3 demonstrates that both structural models fit the data well To test H3 and H7, which relate to spill-over effects, and examine the indirect effect of influencer betrayal on other constructs, we used SPSS AMOS 25 to conduct an indirect effect analysis with 10,000 bootstrapped samples and a 99% confidence level for the confidence intervals; the results of the indirect model show that influencer betrayal has negative indirect effects on influencer coolness (H3: In the brand transgression condition, the results of the structural modelling revealed that brand betrayal has negative effects on brand attitude (H8: β = -.62; t = -6.64, p ≤ .001) but does not significantly predict brand trust (H9: β = .01; t = 0.04, p > .05). As expected, brand attitude significantly and positively predicted brand trust (H5: β = .92; t = 13.35, p ≤ .001), brand trust is positively associated with purchase intention (H6: β = .85; t = 14.39, p ≤ .001), brand trust significantly and positively predicted influencer coolness (H10: β = .19; t = 1.92, p = .054) and, finally, influencer coolness was significantly and positively related to parasocial relationship (H11: β = .78; t = 7.93, p ≤ .001). Thus, H8, H9, H10 and H11 are supported, whereas the results of H5 and H6 were replicated.
In terms of the spill-over effects testing, we followed a similar indirect effect analysis as in the influencer transgression model. To ensure the robustness of our hypothesised models, we further conducted a respecified model by investigating the significance level of the path estimates that were included in the hypothesised models (appendices A and B). Apart from the hypothesised path estimates, there was no additional significant level of path estimates in the respecified model. Thus, our hypothesised model has fulfilled its effect stability requirements and has demonstrated a high degree of internal validity (Hair et al., 2010).

General Discussion
In the current study, we found that influencer betrayal after an influencer transgression scenario negatively affects the follower's parasocial relationship with the influencer, which fully mediates the relationship between the betrayal and influencer coolness. Influencer betrayal was also found to negatively affect brand attitude, brand trust and purchase intention through parasocial relationships and influencer coolness. This suggests that a betrayal by an influencer affects not only the relationship that the influencer has with their own followers but also the brand that the influencer has endorsed and the relationships that individuals have with the brand.
We also found that brand betrayal after a brand transgression scenario negatively affects brand attitude, which fully mediates the relationship between the betrayal, brand trust and purchase intention. Brand trust was also found to be positively associated with influencer coolness. Therefore, brand betrayal was found to negatively affect the coolness of the influencer who had endorsed the brand and parasocial relationships with the influencer through brand attitude and brand trust. This goes to show that endorsing brands that commit moral violations can be damaging to the relationships that influencers have with their followers.  Our research contributes to understanding the dark side of influencer endorsements and how relationships between social media influencers, brands and individuals can also break. Although several earlier studies have examined the influence of negative celebrity information on brands (e.g., Carrillat et al., 2014;Kelly et al. 2016;Till and Shimp 1998;Zhou and Whitla 2013), little research has been carried out on how influencer betrayals affect endorsed brands and, conversely, how brand betrayals affect endorsing influencers. While earlier literature recognises influencer transgressions (Cocker et al., 2021), their dynamics are still understudied. Our findings complement the study of Colliander and Erlandsson (2015), who found no significant negative impact on the part of blogger transgression on brand attitude. In the current study, such a connection was found, not directly but through the construct of influencer coolness.
Another theoretical contribution lies in extending the construct of brand coolness (Warren et al., 2019) to the context of social media influencers. Influencer coolness was found to be positively associated with parasocial relationships and brand attitude. Because brand coolness has been seen as a desirable and profitable feature for brands (Warren and Campbell, 2014), our results suggest that it is a desirable feature for influencers as well. We also demonstrated both the vulnerability of influencer coolness after influencer and brand betrayals and the importance of coolness in brand endorsements through the connection that coolness has with brand attitude and brand trust.
The current study also extends the literature on brand betrayal (Tan et al., 2021) to the context of influencer betrayal. Moral violations on the part of brands can affect the relationships that individuals have with brands (Reimann et al., 2018), and the feelings of betrayal caused by influencers can affect their relationships with their followers. A further finding is that these emotions can spill over, even to seemingly innocent parties: although, in our research design, the condition of the influencer transgression scenario was unrelated to the associating brands and the brand transgression scenario was unrelated to the endorsing influencers, the effect can still be traced to brands and influencers, respectively. This indicates that partners can be found 'guilty by association' (Carrillat et al., 2014), even though they are innocent of the transgression made by the other party and may not even be aware of it. However, the results may also indicate 'double betrayal', in which an individual first feels betrayed by a favoured brand and is then hit by another wave of betrayal after realising that a favoured influencer has endorsed such a deceitful and unethical brand. The individual has been led on by not just one but two trusted friends.
In terms of methodology, our research contributes to instrument development (i.e., measurement items for influencer betrayal), emphasizing how to tackle a sequential negative response from consumers or followers after an influencer transgression.

Practical implications
Collaborations between social media influencers and brands are becoming more common, and long-term relationships are becoming the ideal, instead of short campaigns (Influencer Marketing Hub, 2020). However, as the relationships between influencers, followers and brands become intertwined, both positive and negative emotions can become entangled. Therefore, it is important to understand the psychology behind these relationships.
If negative associations spill over from the influencer to the brand and from the brand to the influencer, both parties should weigh the potential benefits and consider the potential losses caused by a collaboration. Strategic cautiousness and the probing of partnerships may prove valuable. In fact, the mere awareness and acknowledgement of a potential spill over may be beneficial because such events may occur even when the parties are unaware of these links (Bowden et al., 2017).
Brands have been understood as increasingly responsible for their actions' effects of society at large, and our findings call for a new type of responsibility: collaboration responsibility. Collaboration responsibility is expected from both brands and influencers in that existing associations or past transgressions should be openly discussed before collaborations begin. This responsibility implies a shift from the typical self-care approach, in which the parties are mostly worried about their own reputation, toward an understanding that influencers and brands have responsibilities regarding one another's reputation. This extended responsibility should reach the influencers' followers, brand customers and other stakeholders and include considering how these other parties may be affected and the potential long-term implications of Notes: ** p ≤ .001; * p ≤ .05; *n p = .054 CI = confidence intervals; ª Not included in the analysis; Indirect test with 10,000 bootstrapping and a 99% confidence level for confidence intervals collaboration. For brands, collaboration responsibility also includes ensuring that the cooperation allows the influencer to maintain or even deepen the relationship the influencer has with his/her followers. Moreover, because many influencers and their followers are still very young and adolescents are known to both become elated by brands but also regret their commitments (Dhir et al., 2016: 427), brands should be held to a greater standard in terms of ensuring the safety and well-being of young people.

Limitations and future research
The present research has certain limitations, but it also opens avenues for additional studies. First, a real brand and real influencer were used in this research, and the original attitude towards the influencer and brand may have affected the proposed effect. In fact, all experiments carry their own risks, and in the spirit of collaboration responsibility, we are aware of the risk that our conditioning could have formed unwanted associations in the minds of the respondents. To advance our understanding of the field, however, we feel that the use of self-reported influencers and brands was necessary. In the future, a fictitious influencer and hypothetical brand could be used to test the effect. The use of online experiments with MTurk also raises the issue of validity. Replication in a field setting with a larger sample size might be helpful in future studies; in addition, more culturally and geographically diverse responses would help as well.
The research design contained two conditions with two transgression scenarios. A control group with no transgression scenarios would have helped to verify the results. The results were also limited to the presented scenarios. The precise scenarios were chosen to test the model, but other potential transgressions should be tested in future studies. Moreover, future studies should measure the severity of various kinds of transgressions and the margin for operations on both sides regarding different transgressions. We also acknowledge that the spill-over effect may be offset by other potential moderating effects, such as self-brand connection or self-congruency (Tan et al., 2019) with the social media influencer. Thus, future studies should test the model with additional moderators.

Conclusion
The current study explores the intertwined relationships between social media influencers, brands and individuals on social media. Although positive emotions have been found to spill over from influencers to brands (Reinikainen et al., 2020), the effects of negative emotions have mostly been overlooked, such as the negative feelings that are caused by trust violations. The present study specifically examined the effects of influencer and brand betrayals on endorsed brands and endorsing influencers, respectively.
The results show a risky two-way effect: a transgression on the part of either party can reflect on the other party by shattering trusting relationships. This is something that both influencers (Lövheim, 2011) and brands (Winchel, 2018) contemplate, and the current study shows that the risk is real. Specifically, we demonstrate that an influencer transgression has resulted in negative impacts on the endorsed brand's attitude, trust and purchase intention, whereas a brand transgression would negatively affect the endorsing influencer coolness and parasocial relationships with their supporters.
Based on the findings, one might ask whether breaks in trust in the relationships between influencers, followers and brands could hold a deeper significance for society at large. Individuals' experiences with other individuals, institutions and organisations have been linked to the level of generalised trust in society (Rothstein and Uslaner, 2005). It has also been suggested that increased levels of distrust lead to more distrust, thus creating a 'vicious circle of distrust' (Canel and Luoma-aho, 2019). Whether this same logic applies to influencers and brands remains unclear, but it should be considered a possibility that these feelings of betrayal could spill over to other areas of individuals' lives, thus diminishing the likelihood of trusting others and even contributing to diminished generalised trust in society.