2.1 Influencers

Influencers are individuals who create valuable content, have high reputations in specific fields (Cha et al., 2010; S. Kim et al., 2017) and are followed by a large number of users on online social networks (De Veirman et al., 2017).

As influencer marketing has become more popular in recent years, scholars have investigated its success factors. Wiedmann and von Mettenheim (2018, 2020) presented an overview of the success factors of endorsers and influencers in particular and suggested five main categories of success factors: (a) endorser distinctive factors, which describe factors that are inherent to the endorser (e.g., attractiveness, trustworthiness); (b) perceiver congruence factors, which involve the interplay of the endorser with a targeted audience; (c) brand/product congruence factors, which describe the interplay of the endorser with the brand; (d) management factors, which include “behind-the-scenes” administrative issues such as the financial constraints of the endorsement and, finally, (e) communication factors, which are related to the issue of whether an endorser can communicate in an adequate way, for example, the suitability of his or her voice. Given the goals of the current work, the literature review will focus especially on literature addressing the categories of (a) perceiver congruence factors and (b) brand/product congruence factors. Influencer can be classified based on their amount of followers and influence as either mega-influencers, macro-influencers, micro-influencers or nano-influencers. While the “bigger” types of influencers resemble celebrities, the smaller types of influencers (Geyser, 2017), on which the focus lies in the course of this research differ from celebrities by being perceived as to be more authentic, closer to consumers and provide a more interactive communication experience (Djafarova & Rushworth, 2017). Moreover, issues regarding brand congruence has been outlined to be of particularly high importance for them as the endorsement of appropriate brands is part of their self-conception (Geyser, 2017) .

The congruence of the influencer with the following constructs will be the subject of this research:

Brand personality is “the set of human characteristics associated with a brand” (Aaker, 1997, p. 347). It embodies knowledge and shapes brand perceptions (Freling & Forbes, 2005). Models conceptualizing brand personalities similar to those of humans have been developed (e.g., Aaker, 1997). Brand personality is regularly used as a vehicle to assess how similar (or dissimilar) a brand is to another entity (a new product category, another brand, an event, or an individual) (Fleck & Quester, 2007; Maille & Fleck, 2011). In this way, it also appears to be well suited for a comparison between a brand and an influencer.

The actual self is defined as the authentic self, which is related to who an individual is at present. Hollenbeck and Kaikati (2012) emphasized that the need for self-congruity accrues from the motivation to “[maintain] the coherence of a personal conceptual system” (Epstein, 1992). Once such a set of beliefs is established, individuals strive to maintain it (Klipfel et al., 2014). The ideal self, in contrast, is defined as the individual's idea of how he or she would like to be (Dolich, 1969). It is shaped by an individual's imagination of ideals and aspirational goals (Lazzari et al., 1978; Wylie, 1979).

2.2 Current research situation

In the extant research on (a) perceiver congruence factors, many questions remain unanswered. While there is a tendency to affirm that the actual self-congruence between a user and an influencer can positively affect the influencer and the brand (e.g., Shan et al., 2020; Sokolova & Kefi, 2020), the relevance of the desired self remains much more obscure. At best, the findings of Schouten et al.'s (2019) comparative study shed some light on this issue. Against their hypothesis, they found that influencer endorsements led to more wishful identification than celebrity endorsements. However, a study examining the desired self-congruence of an influencer as an independent variable and its impacts on influencer and brand-related constructs is still missing. Concerning the moderating effect of involvement, the current research situation is similarly very scarce. In particular, to date, no study has analyzed the moderating effect of involvement and desired self-congruence in the context of influencer marketing.

Concerning the (b) product-related factors, the extant body of literature has obtained conflicting results on whether influencer-product congruence is beneficial or unimportant (Breves et al., 2019; De Cicco et al., 2020; D. Y. Kim & Kim, 2020), and it has pinpointed the need for further research on this issue. Moreover, no study has examined the potential moderating effect of involvement in the context of influencer-product congruence.

In conclusion, the following key gaps in the research can be identified: The impacts of ideal self-congruence and product endorser-congruence in particular appear to be underexplored. Similarly, the moderating role of involvement on the different types of congruence has been very scarcely considered. It should be stressed that adapting findings developed for celebrities does not appear to be a solution in this context as the differences in the action mechanisms of the variables are not yet well understood and are often found to be very remote from that suggested by theoretical considerations and researchers' intuition (Schouten et al., 2019; Trivedi & Sama, 2019; Xiao et al., 2018).

2.3 Basic theories

The relevance and methods of operation of three forms of congruence - (1) the brand, (2) the actual self or (3) the ideal self - are explained by two theories: (1) Kelman's (1961) theory on opinion change and (2) Kahle et al.'s (1986) social adaptation theory. Involvement is conceptualized by (3) Petty et al.'s (1981) elaboration likelihood model.

These theories have been found to work well together (e.g., Petty & Cacioppo, 1986) and have also been used simultaneously in the prior research (e.g., Y. Lee & Koo, 2016).

2.4 Theory on opinion change

The theory on opinion change explains the impacts of the actual self and the ideal self. It is a fundamental theory of opinion formation and is designed to help investigators identify the motivations that underlie opinion-changing processes. The effect of brand endorsement on advertising effectiveness is determined by identification (with the endorser). When consumers believe that they share interests, values, or characteristics with an endorser, they are more likely to adopt the endorser's beliefs, attitudes, and behaviors. This belief can accrue either from congruence with the actual self, which is the degree to which individuals perceive that they have commonalities with another individual, or from congruence with the ideal self, which is the desire to be like another individual (Basil, 1996; Kelman, 1961).

2.5 Social adaptation theory

Social adaptation theory illustrates the remaining form of congruence, specifically, the congruence between endorser and brand. This implies that the adaptive significance of information will determine its impact. The processing of information is based on the usefulness of adaptation. If a perceiver finds that a particular source of information does not facilitate adaptation, he or she will stop processing that source (Kahle & Homer, 1985). Based on this reasoning, Kamins (1990) demonstrated that the physical attractiveness of a celebrity endorser positively affects consumers' evaluations of a brand used to enhance one's attractiveness but is of no use if a brand's product has no relationship to physical attractiveness. In general, it can be assumed that when endorsers exhibit any type of high brand congruence, a high level of expertise and credibility is assumed by perceivers (Dwivedi & Johnson, 2013; Y. Lee & Koo, 2015).

2.6 Elaboration likelihood model

The elaboration likelihood model explains the interplay of involvement with the three forms of congruence. It is based on two basic assumptions: (1) People are motivated to hold correct attitudes. (2) Although people want to hold correct attitudes, the amount and nature of issue-relevant elaboration in which people are willing or able to engage to evaluate a message vary with individual and situational factors.

The amount of cognitive processing performed for an attitude change depends on the involvement. Attitude changes occur through two routes: a peripheral route that minimizes cognitive processing and a central route that requires intense processing (Petty & Cacioppo, 1986).

Low involvement occurs when the interest in a stimulus is low (Antil, 1984). The importance of persuasive arguments is small while superficial characteristics are important (Holzwarth et al., 2006; Petty et al., 1981; Petty & Cacioppo, 1979). Attitude change travels through the peripheral route along with simple cues associated with the issue (Roozen & Claeys, 2010). In contrast, in high-involvement conditions, consumers search for information more intensively (Coulter et al., 2003). They are devoted to learning about the true merits of a product and exert the necessary cognitive effort to process issue-relevant arguments (Petty et al., 1983). Elaboration becomes more likely. In this case, the attitude travels through the central route whereby a person exercises diligent consideration of the information (Roozen & Claeys, 2010). The research in cognitive and social psychology provides strong support for the view that, sometimes, people engage in “controlled,” “deep,” “systematic,” and/or “effortful” analyses of stimuli, and, other times, their analyses are better characterized as “automatic,” “shallow,” “heuristic,” and/or “mindless” (Petty & Cacioppo, 1986).

2.7 Theory integration

The elaboration likelihood model can serve as a bracket that helps to tie together Kelman's (1961) theory on opinion change and (2) Kahle et al.'s (1986) social adaptation theory. Petty et al. (1981) suggested that many theories of attitude change could be roughly placed along the elaboration continuum. In their 1986 work, Petty and Cacioppo discuss their elaboration likelihood model with numerous other theories. They thereby convey information on how to harmoniously integrate other theories into their elaboration likelihood model. They provide the general statement that many other theories consider either only (1) low involvement situations, where attitude change travels along a peripheral route or (2) high-involvement conditions, where attitude travels along a central route. Therefore, when integrating a theory in the elaboration likelihood model, it must be checked for whether it is based on (1) low or (2) high involvement conditions. Petty and Cacioppo (1986) explicitly mention Kelman's (1961) theory on opinion change and classify it as a theory whose assumptions and conditions describe a process of attitude change driven by simple affective cues. Therefore, it would operate under low involvement conditions. Petty and Cacioppo (1986) do not refer to Kahle et al.'s (1986) social adaptation theory (probably because both works were published in the same year). Therefore, based on Petty and Cacioppo's (1986) general statement, in the course of hypothesis development, it will be explored whether the theory classifies as either superficial, symbolic information with no data on the true merits of the product or as argumentative, evidence-based information. An important addition to this issue is provided by the works of Y. Lee and Koo (2016) and Handriana and Wisandiko (2017), which forge a link between the elaboration likelihood model and social adaptation theory. For example, Y. Lee and Koo's (2016) work on celebrity endorsement uses both theories to infer that product endorser congruence on the level of expertise and physical attractiveness is more intensively processed under high-involvement conditions.

3.1 The effects of congruence with the brand, the actual self and the desired self

The endorsement of the meat deliverer “Butcher Box” by the vegan influencer Alyse Parker angered the brand's fans. They stated that they found the endorsement incongruous and ridiculous (Parker, 2020). The reasons for this can be found in social adaptation theory, which states that the effectiveness of an endorsement is tied to the degree to which the image, personality, or expertise of the endorser fits the advertised product or brand (Basil, 1996; Kelman, 1961). Even a simple match between the physical characteristics of spokespersons and the perceived characteristics of brands produces effects in product evaluations (d'Astous & Bitz, 1995; Kanungo & Pang, 1973; McSweeney & Bierley, 1984).

Consumers utilize a source of information only to the extent that it facilitates adaptation to environmental conditions. If there is a match between endorser and brand, the endorser becomes an effective source of information with regard to the effectiveness or benefits of the brand (Kamins, 1990). However, if congruence is lacking, unfavorable product evaluations will result because consumers must change their cognitive structures (Kanungo & Pang, 1973).

Numerous scholars have argued that celebrity-brand congruence is a determinant of endorsement effectiveness (e.g., Till et al., 2008; Till & Busler, 1998, 2000). Notably, Choi and Rifon (2012) argued that congruity enhances ad attitude while Kahle and Homer (1985) showed that congruity increases the trustworthiness of communication. However, the process of how and to what extent these findings transfer to influencers is unknown. In theory, there are two opposing schools of thought arguing either for a very high or very low importance relative to other endorser types (Breves et al., 2019). Proponents of a very high importance of congruence argue that influencer marketing is experienced as regular communication without (pure) persuasive intentions. If media users notice a mismatch between the brand and the influencer, they might cognitively stumble over the unsuitable affiliation and consequently perceive the influencer and his or her message to be less credible. They are likely to assume a persuasive and commercial intent as they feel that the influencer wants to palm off the product on them (Evans et al., 2017; Koernig & Boyd, 2009). Therefore, congruence between influencers and brands would be of very high importance. Opponents of this line of thought argue that brand congruence is overshadowed by interpersonal connection. As media users perceive an influencer as one of them, they will seek highly personal advice. As long as this is provided, an actual connection between the influencer and the brand will be irrelevant (Breves et al., 2019).

In the empirical research, Schouten et al. (2019) hypothesized that this type of congruence will be more pronounced for influencers than celebrities. This was presumed to be the case because influencers are viewed as representative of particular domains of interest, such as “beauty vloggers,” while celebrities will not have developed such a distinct, exclusive specialty (Balog et al., 2008; Schouten et al., 2019) However, Schouten et al. (2019) could not confirm this hypothesis.

Overall, it must be noted that there are high theoretical discrepancies regarding the relevance of brand congruence. To make matters more complex, empirical results have been shown to work in a different way than scholars have predicted based on theory, which highlights the relevance of further investigating this issue.

The endorsement of the meat delivery service “Butcher Box” by the vegan influencer Aylse Parker also incensed her followers. Her vegan followers stated that they could no longer identify with her (Parker, 2020).

This finding may be explained by the theory on opinion change. It states that a person who identifies with an endorser is more likely to adopt modeled behaviors and to engage in advocated behaviors (Basil, 1996). Individuals like similar sources more than dissimilar ones (Byrne, 1971). This preference facilitates the flow of information as perceived communication barriers are lower and communication volume becomes higher. Individuals also feel more comfortable choosing a similar source due to presumed common needs (Lazarsfeld & Merton, 1954). Additionally, bearing in mind that consumers use brands to signal their identity and reaffirm their self-image (Bodner & Prelec, 2005; Dunning, 2005), actual self-congruence with the endorser facilitates perceivers to adopt their (positive) perception of the brand.

In the event of actual self-congruence, the persuasiveness of a celebrity endorser increases (Kamins & Gupta, 1994; Pradhan et al., 2014). This should be even more true for influencers. In contrast to celebrities, influencers are perceived as people “like you and me” (Kamps & Schetter, 2018; Wiedmann et al., 2010). In Djafarova and Rushworth's (2017) qualitative interviews on the differences between celebrities and influencers, the participants expressed that it was highly important that an influencer was similar to them (e.g., in terms of personal taste, income or any other reference point). This was not true for celebrities who were perceived as aloof individuals who inhabit another world. This finding was reflected by Shan et al. (2020) who argue that the extent of consumers' actual self-congruence with an influencer leads to a more positive attitude toward brand content.

As research on the effect of similarity for genuine influencers is relatively scarce, further insight can be gained by considering other types of online endorsers: Balabanis and Chatzopoulou (2019) argued that a blogger who was perceived as similar to the information seeker was more influential. Electronic word of mouth (EWOM) stemming from demographically similar sources is more influential than information from dissimilar sources (Steffes & Burgee, 2009). Similarity is crucial in determining credibility perceptions and attitudes toward user-generated content (Ayeh et al., 2013). Of course, it cannot be definitively stated whether these findings developed for other, smaller types of online endorsers apply to influencers with the same strength (Schach, 2018).

Alyse Parker's endorsement of “Butcher Box” enraged a third group of followers. These were the followers striving toward a vegan diet. They expressed that they could no longer admire her as their role model and were highly disappointed (Parker, 2020). In light of these findings, the question arises as to the role played by identity in the context of an influencer endorsement. According to the theory on opinion change, individuals also identify with models that fit their perception of how they would like to be. These models are defined as an actual or imaginary individual conceived as having significant relevance upon an individual's evaluations, aspirations, or behavior (Park & Lessig, 1977). Hence, in an attempt to achieve their ideal self-image, consumers tend to conform to attitudes and behaviors if an endorser's image is congruent with their ideal self-image (Choi & Rifon, 2012).

Empirical findings show that consumers reject brands endorsed by celebrities who do not match their ideal self (Escalas & Bettman, 2017). Congruence with the desired self induces favorable responses to an advertisement (Choi & Rifon, 2012) and positively impacts ad attitude (Çakır & Çakır, 2015). It is, however, not guaranteed that these findings can be adapted to influencers. Celebrities are generally labeled as representing an aspirational reference group for consumers (Dwivedi et al., 2014). They are perceived as highly superficial individuals who exist on a level that consumers would like to reach but cannot actually do so (Temperley & Tangen, 2006). In contrast, consumers perceive influencers as being closer to themselves, less superficial and more down-to-earth, endowing them with great powers of persuasion (Djafarova & Rushworth, 2017). Djafarova and Rushworth (2017) expressed that the promotion of fitness DVDs to address weight-related problems was more persuasive if carried out by influencers than by celebrities. They argued that an influencer was likely to be perceived as having been overweight while a celebrity was considered to be above such ordinary problems. These findings, however, are refuted by Schouten et al. (2019), who found that desired self-congruence was even more important for influencers than celebrities. In light of these conflicting results, the verification of the relevance of ideal self-congruence appears to be a matter of high relevance. Therefore, the following hypothesis is proposed.

3.2 The moderating effect of involvement on congruence with the brand, actual self-congruence and ideal self-congruence according to the elaboration likelihood model

The elaboration likelihood model suggests that, under high-involvement conditions, strong arguments offer more cues to remember than weak arguments and are thus more persuasive. In contrast, under low-involvement conditions, peripheral cues such as admiration of the source are likely to have great impact on persuasion regardless of the argument's strength. Because in high-involvement situations, individuals are more motivated to devote cognitive resources to the cognitively taxing and incremental process of assessing an endorsement of a brand, they pay attention to the quality of an argument and make inferences about the relationship between the brand and its endorser (Johar & Sirgy, 1991; J. G. Lee & Park, 2014; Sirgy & Su, 2000).

Empirically, it has been found that the impacts of endorser-product match develop in a stronger way when consumers are motivated and able to elaborate on information (J. G. Lee & Park, 2014; Peracchio & Tybout, 1996). These findings suggest that the congruence between brand and endorser requires a high amount of cognitive processing so that persuasion travels through the direct route. Therefore, the following hypothesis is proposed.

In their discussion, Petty and Cacioppo (1986) indicated that actual and ideal self-congruence might be more relevant under low rather than high involvement conditions. Although this is an issue that has been acknowledged to be theoretically relevant, relatively little consideration has been given to it. It has been generally stated that studies exploring the interplay of involvement and personality issues are limited (Ekstam & Bjurling, 2018). Fleck et al. (2012), for example, have mentioned that it might be relevant to refine their results on the questions of congruence of celebrity endorsers for different involvement conditions. The role of involvement might even vary among types of social media endorsements. For example, under high involvement conditions, endorsements by influencers have been found to generate higher brand attitudes than other forms of social media endorsements (Ekstam & Bjurling, 2018). It is thus apparent that involvement is a variable that varies individually, and the question of how it interacts with the three types of congruence in influencer marketing is relevant.

Johar and Sirgy (1991) used the elaboration likelihood model to introduce their concept of the “self-congruity route.” The self-congruity route to persuasion describes a psychological process in which consumers focus on source cues and match those cues to their actual and/or ideal self-concept. This route is employed when the involvement level is low as actual/ideal self-congruity classifies as lowly cognitively taxing, holistic and simplistic criteria (Johar & Sirgy, 1991). However, there are also examples of contradictions to this theory. Under high-involvement conditions, consumers tend to rely more on cues; require more information and, in general, think harder. Therefore, it is possible that consumers also process the forms of the actual self and the ideal self in a stronger way (Racherla et al., 2012). In the context of online product reviews, Racherla et al. (2012) found that the effect of perceived similarity between consumer and reviewer was even greater under high involvement conditions.

We are now faced with the contradictory scenario of the findings of Johar and Sirgy (1991) and those of Racherla et al. (2012). Against this backdrop, it has to be discussed which of these studies developed the most pertinent results. It becomes apparent that while Johar and Sirgy (1991) consider multiple variables, Racherla et al. (2012) merely focus on manipulating the one and only variable actual self-congruence. Against this backdrop, it can be assumed, that Racherla et al.'s (2012) participants only processed actual self-congruence in a stronger way because the other relevant information that would normally have been processed under high involvement conditions were missing. Perhaps actual self-congruence was used as a proxy to guess this information. Hence, based on this discussion, it can be supposed the evaluation of the actual and ideal self-congruence of the endorser does not require much cognitive processing and is therefore based on symbolic information. This contradictory situation renders an investigation into this issue highly interesting, and to verify our assumption, we propose the following hypotheses.

3.3 From post perception to brand behavior

Was the anger of the fans of “Butcher Box” or Alyse Parker a rather superficial occurrence, affecting mainly the single endorsement, or did it have deeper, more lasting consequences for the brand?

In reviewing the literature reviews of influencer endorsements in the context of congruence issues, it becomes apparent that most studies have focused on the impacts on the perception of the influencer or the post. In contrast, the impacts of influencers on brand trust and purchase intention have been found to be underexplored (Hermanda et al., 2019; Jiménez-Castillo & Sánchez-Fernández, 2019; Kolarova, 2018). However, for the brand's decision regarding the endorsement, the tangible consequences of that decision are also a matter of high relevance (Jin & Ryu, 2020). In particular, purchase intention can be understood as a widely used marketing tool to estimate the effectiveness of a marketing strategy, which can be used to predict sales and market share (Morwitz, 2012). Therefore, we go a step further in exploring the effects on brand trust and purchase intention.

Trust in a brand can be built through engagement and relationships with the brand (Habibi et al., 2014); however, trust can also be transferred. Trust transfer occurs when initial trust in a target (a person, a group, or an organization) turns into trust in another target (Stewart, 2003). For example, consumers' trust in another consumer's communication in a social media brand community can be transferred to trust in an associated brand (Liu et al., 2018). Trust can alter the favorableness of consumers' opinions and increase the perceived trustworthiness of the endorsed brand (F. Li & Miniard, 2006). This suggests that trust in a social media influencer could also transfer to a brand that the influencer uses or recommends (Reinikainen et al., 2020).

Interestingly, the fundamental question of whether purchase intention can be impacted by influencers is controversial. Some scholars affirm this possibility (Lisichkova & Othman, 2017; McCormick, 2016) while others negate it (Hermanda et al., 2019).

To reconcile these positions, it has been supposed that influencers generally do not directly influence purchase intention; however, there could be an indirect effect through perceptional or behavioral variables (Jamil & Rameez ul Hassan, 2014; Johansen & Guldvik, 2017). Therefore, it can be suggested that brand trust could work as a variable impacting purchase intention as brand trust can be a building block for purchase intention (Dodds et al., 1991).

4.1 Analyzed brands: Nike and Mercedes

To select suitable brands, a pretest (n = 30) was conducted. In the course of the pretest, the participants assessed eight brand scenarios for the perceived level of involvement on a six-item scale adapted from Laurent and Kapferer (1985). All brands were famous throughout Germany where the present research was performed (Junker, 2018). This factor is an important prerequisite because perceptions of brand personality traits are formed by a consumer's prior contact with a brand (Plummer, 1985). The results revealed that the involvement level differed the most and with the highest significance level for the two following scenarios: A Nike unisex sports bag as a low-involvement product and a Mercedes S Class as a high-involvement product (Nike_involvement: = 1.980, SE_Nike = 0.084, Mercedes_Involvement = 3.997, SE_Mercedes = 0.069, p < .05, t = 2.101). A subsequent variance analysis comparing the brands by means of Mäder's (2004) personality scale identified that significant differences between the brand personality of Mercedes and Nike were perceived (p < .01).

4.2 Pretest and stimulus material

The stimulus material consisted of an influencer profile including a post about the endorsed product. To avoid legal issues, profile pictures of existing influencers were simulated with images of similar-looking individuals from image databases. These pictures were completed with the characteristics of the influencers (e.g., field of interest, life motto) inspired by the real role models.

The investigation required influencers to fit well (badly) with the analyzed brands Mercedes and Nike. To select an appropriate set of influencers, a pretest was conducted (n = 30), in which the participants assessed the personalities of the Mercedes and Nike brands as well as a set of 12 influencers on a scale adapted from Mäder (2004). The scale includes five personality constructs: “Attractiveness,” “Reliability,” “Temperament,” “Stability” and “Naturalness.” This scale was explicitly developed to measure the personality of a brand and an endorser, offering an advantage for the present study (e.g., the scale of Aaker, 1997).

To assess the fit of the brands and influencers, the squared Euclidian distance of the perceived differences of all five personality constructs was calculated (Pradhan et al., 2014). The pretest revealed the following results: the well-fitting endorser of Nike (Nr. 3, inspired by the real influencer “Chick‘N'Kicks”) was a sporty-looking young woman. Her life motto was “good vibes, good kicks and power.” The well-fitting endorser of Mercedes (Nr. 1, inspired by the real influencer “Grey Fox”) was an elegantly dressed gentleman. His field of interest was high-quality products of refined design and fine artisanship. Finally, influencer Nr. 2 (inspired by the real influencer “Hawtchocolate Christina”), who was predetermined to be a bad fit for Mercedes and Nike, was a shy-looking young woman whose main interest was food products, especially chocolate. (The stimuli can be found in Figure 1.)

4.3 Survey design

The study employed an online survey with five-point scales to measure the answers of the subjects. Overall, eight variables were considered: Brand Congruence, Actual Self-congruence, Ideal Self-congruence, Involvement, Post Attitude, Post Belief and change in Brand Trust and Purchase Intention. Change was explicitly selected to avoid skewing the results by prior attitudes/intentions or a subjective previously developed desire for the concrete product. In the context of the Mercedes brand, a hypothetical purchase intention (“If I could afford…”) was applied as not all of the participants had the necessary solvency. Actual and Ideal self-congruence were measured on one-item scales adapted from Reed (2004). Involvement was measured with a six-item scale adapted from Laurent and Kapferer (1985). A four-item scale adapted from Aaker (2000) was used to measure post attitude. Beltramini's (1982) six-item scale was employed to measure post belief. Finally, two-item scales adapted from Wiedmann et al. (2014) were used for Brand Trust and Purchase Intention.

4.4 Data collection and analysis

The data collection occurred in Germany via a randomized online survey shared on the popular German research platforms SurveyCircle, PollPool and Thesius as well as among students of German universities from May through September 2019. Only participants who stated that they knew and followed at least one social influencer were eligible for participation. Data on 605 participants were collected. After running the rigorous algorithm Time_RSI, which detects invalid answers by means of the criteria of speed and consistency (Leiner, 2013), valid data from 547 participants (65.3% female, average age: 25 years) were employed. The age distribution was as follows: Age_18-20: 12.6%, Age_21-25: 54.9%, Age_26-30: 24.3%, Age_31-35: 3.3%, Age_{older than 35:} 4.7%. The distributions of occupations were as follows: Student: 58.9%, Employee, 31.0%, Self-employed: 4.3%, Retiree: 1.2%, Other: 4.6%. The relatively young average age and the higher relative proportion of females may be rooted in the fact that influencer marketing appeals more to the younger generation and to women (Nirschl & Steinberg, 2018).

Manipulation checks were carried out by means of variance analysis in SPSS. To reveal the relationships between the variables, we then built a reflective structural equation model in SmartPLS. PLS SEM was appropriate due to its ability to solve the entire system of equations simultaneously through iteration using maximum likelihood (ML) rather than estimating the parameters of each equation independently (Hayes et al., 2017). This consisted of an obvious advantage in light of the complexity of our model involving numerous sequential constructs. Moreover, SEM also has the non-negligible advantage of accounting for random measurement error when estimating relevant effects involving latent variables (Hayes et al., 2017).

4.5 Theoretical model

We examine and build on our conceptual framework (Figure 2a–c). We assess the effect of brand congruence, actual self-congruence and ideal self-congruence on brand trust in analysis 1 (Figure 2a). In analysis 2, we examine the effect of brand congruence, actual self-congruence and ideal self-congruence on the intermediary variables post attitude and post belief and their impact on brand trust (Figure 2b). In analysis 3, we extend our conceptual model (Figure 2c) to examine the moderating role of involvement on the relationship among brand congruence, actual self-congruence and ideal self-congruence on the one side and post attitude and belief on the other side.

4.6 Manipulation checks

In the course of the questionnaire, the participants were randomly assigned (1) to a group with either high or low personality congruence between influencer and brand and (2) to a high or low involvement group.

To verify whether the manipulation of the stimulus material in terms of (1) personality congruence between the influencer and the brand as well as (2) involvement level was perceived as intended, manipulation checks were performed using ANOVAs. The ANOVA on personality congruence between influencer and the brand (N_{Low_Congruence} = 280, N_{High_Congruence} = 267) ascertained a significant difference among the compared groups (M_{Low_Congruence} = 1.757, SE_{Low_Congruence} = 0.060, M_{High_Congruence} = 4.075, SE_{High_Congruence} = 0.059, p < .0001, t = 455.339). Similarly, the ANOVA on involvement (N_{Low Involvement} = 282, N_{High Involvement} = 265) produced significant results (M_{Low Involvement} = 2.358, SE_{Low Involvement} = 0.051, M_{High Involvment} = 3.081, SE_{High Involvement} = 0.060, p < .0001, t = 152.149).

5.1 Model validation

6.1 Theoretical implications

In this study, the effects of influencers' congruence with a brand as well as the actual and ideal self of the consumer on post attitude and believability were investigated. The moderating effects of involvement were recorded. In addition, the subsequent effects of post attitude on post believability on brand trust as well as the effect of brand trust on purchase intention were substantiated.

In line with social adaptation theory, brand congruence was found to have a significant positive effect on post attitude and belief. The effect of brand congruence is the strongest compared to those of the other two types of congruencies. In this way, a controversial question of influencer marketing has been answered. Breves et al. (2019) outlined that congruence between brand and influencer might be either of very high or very low importance for the influencer. Our research is in line with the arguments of the proponents of very high importance, stating that any mismatch of influencer and brand would heavily disrupt viewers' trust as they would assume a purely commercial endorsement motive (Evans et al., 2017; Koernig & Boyd, 2009).

In contrast to what the theory on opinion change had predicted, actual self-congruence was not found to have an effect. Due to a relative lack of prior research on this issue, this hypothesis was also based on other forms of online endorsements such as user-generated content; however, the findings of Schach (2018) should be considered as this work warned that the mechanisms used to derive these findings might not be adaptable to influencers. In fact, influencers can be seen as closer to celebrities, for whom congruence with the ideal self, not with the actual self, is more important. This also aligns with Schouten et al. (2019), who found no significant difference in similarity issues between influencers and celebrities, which, consequently, contradicts Djafarova and Rushworth (2017).

This presumption might also be reflected by the fact that, in line with social adaptation theory, ideal self-congruence was found to have significant relevance. Again, a contribution to clarifying a controversial question was presented. Djafarova and Rushworth (2017) argued that congruence with the ideal self was not expected from influencers. However, Schouten et al. (2019) found that congruence with the ideal self is not only relevant for influencers but even more relevant than for celebrities. Overall, the predictions of the theory on opinion change can only be partially confirmed. These findings, at first glance, stand in conflict with the results of Sokolova and Kefi (2020) who found that actual self-congruence was relevant for influencers. Although they did not control for involvement, it can be assumed that their results are relevant for high-involvement products as they analyzed influencer endorsements for luxury brands. This would be in line with our results as the importance of actual self-congruence was demonstrated to rise with the level of involvement.

The role of involvement in the context of congruence issues was found to be underexplored even for traditional celebrity endorsers (Fleck et al., 2012). Moreover, involvement was supposed to function differently for influencers than for other endorser types (Ekstam & Bjurling, 2018; Trivedi & Sama, 2019). In accordance with the elaboration likelihood model, a positive moderating effect of involvement on brand congruence and a negative effect on ideal self-congruence were found. However, in contrast with the predictions, the moderating effect on actual self-congruence was positive. These findings are in line with a relatively isolated study by Lin and Yeh (2009) on celebrity endorsements. As an explanation/interpretation of these results, it was stated that for high-involvement products, consumers make the cognitive effort to determine what truly suits themselves due to the relatively high financial risk of the investment (Choi et al., 2005; Zhu et al., 2019). Alternatively, the considerations of Racherla et al. (2012) may provide an explanation as they state that under high involvement conditions, consumers generally consider a greater variety of information including actual self-congruence.

Finally, the subsequent effects on brand trust and purchase intention were similarly investigated. In this way, clarification was provided of an issue that has been found to be underexplored (Hermanda et al., 2019; Jiménez-Castillo & Sánchez-Fernández, 2019; Kolarova, 2018) but of high relevance for brands (Jin & Ryu, 2020). The results essentially confirm that attitude toward and trust in a post have an effect on brand trust; brand trust, in turn, can positively impact purchase intention. In this way, it was shown, at least indirectly, that influencers can have an impact on purchase intention.

6.2 Managerial implications

Brand managers continue to struggle with questions of how influencer marketing is defined, what its value is, and how it should be managed. Against this backdrop, they use partially traditional advertising models (e.g., designed for celebrities) (Childers et al., 2019), which obviously produce some limitations with reference to influencer marketing. The findings of this study suggest various strategies that can be effectively employed to enhance consumers' attitudes and trust in influencers' brand-related posts as well as brand trust and purchase intention.

Based on the results of this study, to increase post attitude and post belief, social media managers can be given the following advice: Under low involvement conditions, they should primarily consider congruence with the brand. As a secondary objective, congruence with the ideal self should be envisaged. Conversely, congruence with the actual self does not need to be considered.

However, when the involvement level rises, social media managers should consider influencers' congruence with the actual self while congruence with the ideal self can be given less consideration. Similarly, the impact of congruence with the brand on post belief (but not on post attitude) becomes more important under high involvement conditions.

The findings on the impacts of the three types of congruence on post attitude and post belief should also be acknowledged by influencers when considering whether to accept an endorsement. Influencers may be tempted to accept any endorsement in exchange for an endorsement fee (Breves et al., 2019). However, this study shows that a misfit of certain types of congruence might elicit unfavorable perceptions of their posts.

Furthermore, brand managers should be aware that post attitude and post belief can indeed have a positive impact on brand trust. In turn, brand trust has a positive impact on purchase intention. In this way, brand managers can be reassured that influencers can indeed increase brand trust and purchase intention.

6.3 Limitations and future research

This study was conducted in Germany, that is, in a Western cultural setting. Therefore, specific cultural values might have impacted the results. Research from an eastern cultural perspective in the context of celebrities has suggested that the relevance of actual self and desired self might vary among eastern and western cultural contexts (Zhu et al., 2019). Therefore, in future research, the results of this study could be compared with one performed in an eastern cultural context.

In further research, more than two (extreme) levels of congruence could be employed. Meyers-Levy and Tybout (1989) suggested that a moderate level of incongruence between an expectation and an object could be beneficial if it is perceived as unexpected and interesting. This might even elicit a positive effect on ad and brand attitude (Harmon-Kizer, 2014). Against this backdrop, it could be investigated whether a mismatch may have contradictory effects in the context of influencer marketing.

Our study introduced involvement as a moderating variable. However, it also did not exclude the existence of further moderators of the three types of congruence. For example, long-term bonding might decrease the effect of congruence between influencer and brand (Breves et al., 2019).

Our research has shown that favorable influencer marketing not only impacts post perception but that it also has an impact on brand-related constructs. Future research could go a step further and consider the impacts on the revenues of a brand. In this way, interesting questions such as “What is the financial value of finding an influencer who is more congruent with the brand?” could be answered. To go a step further, it would be expedient to develop an algorithm that assesses the three types of congruence and can suggest influencers based on the brand and the target group.

An important limitation of this study is that we considered only material goods. In future research, service endorsements could also be considered. The extant research on influencers in the context of service marketing has suggested that due to their intangible nature (it is impossible to touch or see a service) and the impossibility of returning it, consumers perceive the consumption of a service as riskier and therefore devote more cognitive processing to its purchase (Meffert et al., 2018). Against the backdrop of our study, whether brand congruence and actual self-congruence are more important for services could be investigated. Moreover, future studies could seek to replicate the reported results using divergent types of influencers, including more diverse brands, and incorporating different settings to generalize the findings.

This study represents consumers' viewpoints on influencer communication. However, to explore the process from a different perspective, it would be worthwhile to also record the opinions of influencers and practitioners on congruence issues. Wiedmann and von Mettenheim (2020) indicated that the perceptions of consumers, influencers and practitioners on the success factors of an endorsement might vary.

Furthermore, the possibility of personality transfer from an influencer to a brand or vice versa could be investigated. As it is possible that the repeated paring of the two subjects could provoke a perceived convergence (Ambroise et al., 2014), brands may “shape” an influencer's personality over time by means of repeated endorsements (or vice versa). It could also be investigated whether congruence could be to some extant staged. The research on celebrities has demonstrated that the perception of congruence with a brand can be impacted by the script of the advertising object (Pringle, 2004; Pringle & Binet, 2005).