How Social Media Influencers Foster Relationships with Followers: The Roles of Source Credibility and Fairness in Parasocial Relationship and Product Interest
社交媒体影响者如何促进与追随者的关系：来源可信度和公平性在超社会关系和产品兴趣中的作用

Parasocial Relationship 超社会关系

Influencers, to a large extent, provide an indirect and mediated communication channel between marketers/brands and followers. What is unique in the indirect communication through influencers is that marketers’ objectives, such as shaping corporate brand perception, are achieved through the interactions between influencers and their followers (Booth and Matic Citation2011). Therefore, understanding the relationship between influencers and followers is imperative. Horton and Wohl (Citation1956) described parasocial interaction as audiences’ illusory social experiences with media personae. The terms parasocial interaction (PSI) and parasocial relationship (PSR) have been used interchangeably in some literature (e.g., Escalas and Bettman Citation2017; Kim and Song Citation2016). PSR is based on PSI and is considered as a socioemotional bond between media personae and audiences (Horton and Wohl Citation1956; Giles Citation2002). PSI refers to audiences’ relationship perception during a one-time exposure to media production (e.g., show, movie), whereas PSR refers to a more lasting relationship between media personae and audiences (Dibble, Hartmann, and Rosaen Citation2016).
在很大程度上，影响者在营销人员/品牌和追随者之间提供了一个间接和中介的沟通渠道。通过影响者进行间接沟通的独特之处在于，营销人员的目标，例如塑造企业品牌认知，是通过影响者与其追随者之间的互动来实现的（Booth and Matic 2011）。因此，了解影响者和追随者之间的关系势在必行。Horton和Wohl（1956）将超社会互动描述为受众与媒体角色的虚幻社交体验。术语超社会互动 （PSI） 和超社会关系 （PSR） 在一些文献中可以互换使用（例如，Escalas 和 Bettman 2017;Kim 和 Song 2016）。PSR 基于 PSI，被认为是媒体角色和受众之间的社会情感纽带（Horton 和 Wohl 1956;贾尔斯 2002 年）。PSI 是指观众在一次性接触媒体制作（例如，节目、电影）期间的关系感知，而 PSR 是指媒体角色与受众之间更持久的关系（Dibble、Hartmann 和 Rosaen 2016）。

Similar to real-life relationship development, PSR is developed via audiences’ reactions to and involvement with media personae (Stever Citation2017). Unlike social relationship, PSR, as a one-sided interaction, does not entail reciprocity between media personae and audiences (Calvert and Richards Citation2014). Compared with traditional media, social media affords two-way interaction opportunities between influencers and their audiences. Therefore, today’s media personae not only engage in one-way communication with their fans via traditional media channels but also actively engage with their audiences through interactive social media channels such as Twitter or Instagram (Bennett Citation2014). Although PSR has often been examined in the traditional TV era (e.g., Auter Citation1992; Rubin Citation2002), recent studies have extended its application to the interactive social media context (e.g., Colliander and Dahlén Citation2011; Tsai and Men Citation2013).
与现实生活中的关系发展类似，PSR是通过受众对媒体角色的反应和参与来发展的（Stever 2017）。与社会关系不同，PSR作为一种单方面的互动，并不意味着媒体角色和受众之间的互惠（Calvert and Richards 2014）。与传统媒体相比，社交媒体为网红和受众提供了双向互动的机会。因此，今天的媒体人不仅通过传统媒体渠道与粉丝进行单向交流，而且还通过Twitter或Instagram等互动社交媒体渠道积极与受众互动（Bennett 2014）。尽管PSR在传统电视时代经常被检查（例如，Auter 1992;Rubin 2002），最近的研究已将其应用扩展到交互式社交媒体环境（例如，Colliander and Dahlén 2011;Tsai and Men 2013）。

Prior literature examined parasocial relationships between a media celebrity/public figure and his or her fans/supporters across different scenarios, including celebrity endorsement (Escalas and Bettman Citation2017; Wen Citation2017; Yuan, Kim, and Kim Citation2016), political voting (Thorson and Rodgers Citation2006), romantic attachment to media figures (Erickson, Harrison, and Cin Citation2018), brand page interaction (Tsai and Men Citation2013), and interaction with game avatars (Jin and Park Citation2009). More recently, multiple studies have examined the parasocial relationships between social media users, fellow users, and/or social media personae/influencers (e.g., Bond Citation2016, Citation2018; Chen Citation2016). Specifically, Bond (Citation2016) indicated that adolescents on Twitter reported stronger parasocial relationships with their favorite personae when they experienced social interactions than those without such interactions. Similarly, Chen (Citation2016) also emphasized the role of social interactions facilitated by YouTube in the development of parasocial relationships between amateur YouTubers and their viewers.
先前的文献研究了媒体名人/公众人物与他或她的粉丝/支持者在不同情况下的超社会关系，包括名人代言（Escalas and Bettman 2017;温 2017;Yuan， Kim， and Kim 2016）、政治投票（Thorson and Rodgers 2006）、对媒体人物的浪漫依恋（Erickson， Harrison， and Cin 2018）、品牌页面互动（Tsai and Men 2013）以及与游戏头像的互动（Jin and Park 2009）。最近，多项研究考察了社交媒体用户、其他用户和/或社交媒体人物/影响者之间的超社会关系（例如，Bond 2016、2018;陈 2016 年）。具体来说，Bond（2016）指出，Twitter上的青少年在经历社交互动时，比没有这种互动的青少年报告了更强的超社会关系。同样，Chen（2016）也强调了YouTube促进的社交互动在业余YouTube用户与其观众之间超社会关系发展中的作用。

In understanding parasocial relationships, researchers have used both uses-and-gratification theory and uncertainty reduction theory to measure interaction. They have also argued that individuals need to be goal directed and actively analyze the media personae and their behavior during the interaction when developing parasocial relationship (Perse and Rubin Citation1989). Although it has been recognized by researchers that both the performer’s own characteristics (such as attractive or not) and their performance during the interaction (such as verbal or bodily behavior), contribute to audience’s parasocial experience with the performer (Hartmann and Goldhoorn Citation2011), current research on parasocial relationship predominantly investigated characteristics of the performer/source and audience (such as level of loneliness). Based on the previous studies, we argue that in the process of developing parasocial relationship, three aspects play significant roles: the characteristics of the source/communicator, the characteristics of the audience, and the interaction process between the two, research that focuses on the role of communication process or interaction process is scarce (for a review, see Bond Citation2016). Given that the interaction between influencers and followers is more interactive than traditional celebrity–viewer interaction, it is imperative to look into the effect of communication process when studying the appeal of influencers among followers.
在理解超社会关系时，研究人员同时使用使用和满足理论和减少不确定性理论来衡量互动。他们还认为，在发展超社会关系时，个人需要以目标为导向，并积极分析媒体角色及其在互动过程中的行为（Perse and Rubin 1989）。尽管研究人员已经认识到，表演者自身的特征（例如是否有吸引力）以及他们在互动中的表现（例如言语或身体行为）都有助于观众与表演者的超社会体验（Hartmann and Goldhoorn 2011），但目前对超社会关系的研究主要调查了表演者/来源和观众的特征（例如孤独程度）。基于以前的研究，我们认为在发展超社会关系的过程中，三个方面起着重要作用：来源/传播者的特征、受众的特征以及两者之间的互动过程，关注沟通过程或互动过程作用的研究很少（有关评论，请参阅Bond 2016）。鉴于网红和追随者之间的互动比传统的名人与观众的互动更具互动性，因此在研究网红在追随者中的吸引力时，必须研究沟通过程的影响。

We argue that two major components that capture the characteristics of message source and communication process—source credibility and justice—determine the development of parasocial relationship between influencers and followers. Specifically, we expect that influencer credibility and interaction fairness determine the strength of the parasocial relationship between influencers and their followers. In the following section, we review the body of literatures on source credibility and justice in relation to the influencer context.
我们认为，捕捉信息来源和传播过程特征的两个主要组成部分——来源可信度和正义性——决定了影响者和追随者之间超社会关系的发展。具体来说，我们预计影响者的可信度和互动公平性决定了影响者与其追随者之间超社会关系的强度。在下一节中，我们将回顾与影响者背景相关的来源可信度和正义性的文献。

Source Credibility of Influencers and Parasocial Relationship

A number of studies have identified the antecedents of the strength of parasocial relationships, such as the time spent with media personae (Schiappa, Allen, and Gregg Citation2007), characteristics of the individuals (Rosaen and Dibble Citation2016), and characteristics of the media personae (Bond Citation2018).

In terms of the characteristics of influencers, previous researchers have considered the credibility of a message source when gauging a communicator’s influence on the effectiveness of persuasive messages (e.g., Giffin Citation1967; Hovland and Weiss Citation1951; McGuire Citation2001). Through the lens of source credibility, previous studies have investigated celebrity endorsers’ influence on consumers (e.g., Dwivedi, Johnson, and McDonald Citation2015; Guido and Peluso Citation2009; Lee and Koo Citation2015). We align with a recent study and focus on the role of influencer source credibility (Lou and Yuan Citation2019) and expect that source credibility will play an important role in parasocial relationship between influencers and followers.

This study also adopts the four-dimension conceptualization of source credibility (Munnukka, Uusitalo, and Toivonen Citation2016) to examine influencer credibility: trustworthiness, expertise, similarity, and attractiveness. Expertise and trustworthiness are the two original determinants that comprise source credibility proposed by researchers (Hovland, Janis, and Kelley 1953). Source expertise refers to the competence or capability of a source such as the person’s expertise/skills in a certain area or subject (McCroskey Citation1966). Source trustworthiness refers to the extent to which a source is perceived as honest, sincere, or truthful (Giffin Citation1967). Meanwhile, attractiveness and similarity have been identified as antecedents of parasocial relationships by previous researchers (Bond Citation2018; Cohen Citation2009). Attractiveness describes the physical or social attractiveness of the individual who serves as the media persona (Schiappa, Allen, and Gregg Citation2007). In a way that is similar to social relationship development, individuals are more likely to develop relationships with media personae who are attractive (Hoffner and Buchanan Citation2005). More importantly, perceived attractiveness also has a positive effect on the quality and intensity of a parasocial relationship (Schmid and Klimmt Citation2011). Last, similarity refers to mutual characteristics that audiences share with media personae (Schiappa, Allen, and Gregg Citation2007). A number of studies have found that perceived similarity also can lead to more positive interpersonal liking (Duck and Barnes Citation1992), which thus contributes to the strength of parasocial relationship.

Previous researchers have focused on the influences of attractiveness and similarity on parasocial relationships (Bond Citation2018), yet there is a gap in the literature with respect to how perceived expertise and trustworthiness affect the parasocial relationship. The effects of source credibility, as reviewed, provide a logical prediction that perceived expertise, trustworthiness, attractiveness, and similarity of the influencer are important factors determining the strength of parasocial relationship between influencers and followers. Therefore, we propose:

Fairness and Parasocial Relationship

Previous research has focused more on how Internet or social media use facilitates the relationship between communicators and audiences and has not fully investigated how the use of social media challenges the “fundamental notions of relationships” between them (Giles Citation2002, p. 285). While source credibility captures the characteristics of communicators, justice is used to access the interaction process between the communicators and their audience. Justice has often been studied in the context of organizational communication in which employees’ perceived justice of the organizations’ handling of them has been found to shape employees’ trust in the organization, job attitudes, and relationship with the organization (e.g., Colquitt et al. Citation2001; DeConinck Citation2010). The terms justice and fairness have been used interchangeably in the literature. The dimensions of justice/fairness have been widely applied in explicating the persuasion process across varied contexts, including science communication (Besley, McComas, and Waks Citation2006), knowledge sharing in virtual communities (Fang and Chiu Citation2010), buyer–supplier relationships (Liu et al. Citation2012), instructor–student interactions (Chory Citation2007), and customer satisfaction (Martínez‐Tur et al. Citation2006). Using these studies as a backdrop, in the current influencer context we argue that justice/fairness can also inform influencer–follower relationship building.

The literature on fairness investigates behaviors that happen during interactions between organizations and the public, or between decision makers and audiences, which focuses on how peoples’ judgment of fairness affects their views about the outcomes, such as whether the respondent is willing to continue interactions with decision makers in the future (Besley, McComas, and Waks Citation2006). Thibaut and Walker (Citation1975) argued that people care not only about the outcome fairness of a decision (“distributive” fairness)—namely, whether they get a fair outcome for their efforts; they also care about the nonoutcome process, which is the process through which a decision is being made (Thibaut and Walker Citation1975). Lind (Citation2001) found that people use their evaluations of the quality of a decision-making process as heuristic cues to evaluate the legitimacy of a decision when they are unsure about what constitutes a right decision.

There are four factors that constitutue communication justice or fairness, including both outcome and nonoutcome components (for a review, see Colquitt Citation2001). The first component, distributive fairness, is an outcome-related perception that describes the degree to which an individual receives a fair outcome from a decision (Thibaut and Walker Citation1975). The second dimension is procedural fairness, a nonoutcome component which focuses on the degree to which a process was seen as procedurally fair, with an emphasis on whether those affected by a decision had a meaningful voice in the decision-making process (van den Bos and van Prooijen Citation2001). The third dimension is interpersonal fairness, also a nonoutcome component which can be understood as the degree to which decision makers treat those affected by decisions with respectful and polite approaches (Bies Citation2005). The last dimension is informational fairness, a nonoutcome component which argues that people have fair access to appropriate information in the decision-making process (Colquitt et al. Citation2001). Among the four dimnsions, only distributive fairness focuses on the outcomes of a decision; the other three emphasize the behaviors associated with the process. In this study, we posit that all four dimensions can be applied to account for influencer–follower interactions—for example, to what extent followers benefit from the content shared by influencers (distributive fairness); to what extent followers have the chance to share their voice with influencers (procedural fairness); to what extent followers are treated respectfully during their interactions with influencers (interpersonal fairness); and to what extent influencers deliver information to followers honestly and ethically (informational fairness). The difference between informational fairness and distributive fairness is that distributive fairness focuses on the usefulness of the information, while informational fairness focuses on the honesty and accuracy of message delivery.

To the best of our knowledge, no study has considered the fairness of the communication process or similar terms as the antecedents of the parasocial relationship between followers and influencers. Prior studies have shown that these four components affect people’s perceptions of the decision makers and/or relationship with the organizations (e.g., Bies Citation2005; Colquitt et al. Citation2001). Researchers also argued that all four dimensions of fairness contribute to decision making related to interpersonal communication outcomes. For instance, Fang and Chiu (Citation2010) found that all four dimensions positively affect users’ trust in members of virtual communities during the information-sharing process. Some research examined the effects of some specific dimensions on the evaluation of communication outcomes. For example, one study showed that individuals found the participatory decision-making approach received more support and satisfaction from those affected in risk communication contexts (Lauber Citation1999). These studies have provided logical support for us to apply these four dimensions of fairness in the current influencer–follower context. Based on these discussions, the rationale for taking into account the effect of fairness in parasocial relationships is that, different from celebrities, social media influencers’ fame and reputation are heavily based on their interactions with their followers over time. It is reasonable to expect that their communication patterns or process as a whole will influence their relationship building and how followers evaluate the parasocial relationships that arise from this process. Specifically, for influencer–follower communication, we expect that followers’ perceived fairness of their interactions with influencers will be positively related to the strength of their parasocial relationships with influencers. Therefore, we predict:

The Mediating Role of Parasocial Relationship in Product Interests

Besides identifying the antecedents of the parasocial relationship between followers and influencers, how the strength of this parasocial relationship influences marketing-related outcomes, such as interests in endorsed brands or purchase intentions, is also worth testing. One important outcome valued by marketers or advertisers is followers’ interests in the products that the influencer has mentioned (Jin and Phua Citation2014).

Despite a large number of studies investigating the effect of celebrity endorsers on advertising (for a review, see Bergkvist and Zhou Citation2016), influencers exhibit unique traits that distinguish themselves from celebrities, such as influencers’ fundamental role of being content generators on social or online media (Lou and Yuan Citation2019). Moreover, although some studies have addressed the effects of influencer advertising (e.g., De Veirman, Cauberghe, and Hudders Citation2017; Djafarova and Rushworth Citation2017), few studies have addressed the fundamental mechanism of the effectiveness of influencer marketing. For instance, one recent study indicated that perceived trust in branded posts is a mechanism that explains followers’ awareness of the promoted brand and purchase intentions (Lou and Yuan Citation2019). Another study suggests blogger characteristics and blog content affect followers’ engagement with influencer ads, which further affects advertising effectiveness (Hughes, Swaminathan, and Brooks Citation2019). In addition, recent research on the appeal of influencers among adolescents found that parasocial relationships mediate influencer credibility and content value but not parental mediation on adolescents’ materialistic views and purchase intentions (Lou and Kim Citation2019).

Collectively, we propose that the strength of a parasocial relationship serves as the underlying mechanism of influencer credibility and communication fairness on followers’ product interests. As mentioned, the four dimensions of influencer source credibility affect the strength of parasocial relationships, and these four dimensions of a communicator’s credibility have been tested by previous researchers to gain desirable communication outcomes (Belch and Belch Citation2004). Therefore, we propose the following hypotheses to test the mediating role of a parasocial relationship between its antecedences and consequence.

Similarly, as discussed earlier, the four dimensions of fairness are expected to contribute to the formation of parasocial relationships. We also predict that the parasocial relationship mediates the effect of perceived fairness of the influencer–follower interaction on followers’ interests in the promoted products:

Sample

Participants of this study were recruited through Amazon’s Mechanic Turk, and the study was administered online via Qualtrics. To ensure that the participants recruited had interacted with social media influencers before, we included five prescreening questions (e.g., asking them to name one influencer whom they like and whom they have been following; those who could not name an influencer were excluded). A total of 799 participants answered the screening questions, and 426 were eligible and completed the survey. After deleting participants who failed to answer the attention-check questions correctly or who listed obvious noninfluencer names (e.g., “Facebook,” “yes”), the study input a total of 355 participants for data analysis. The participants aged from 19 to 75 (M_age = 34, SD = 10.17), 57% female. The majority of the participants were White/Caucasian (79%), followed by Asian (13%) and Black/African American (8%). Over half of the participants held a bachelor’s degree (55%), and roughly one-fourth were high school graduates (26%).

We asked the amount of time that participants spent on social media accounts on a daily basis. Response options were 10 minutes or less (10%), 11 to 30 minutes (23%), 31 to 60 minutes (20%), 1 to 2 hours (22%), 2 to 3 hours (15%), and Over three hours (10%). About 41% of participants were following 1 to 10 influencers, 28% were following 11 to 30 influencers, 12% were following 31 to 50 influencers, and 18% were following over 50 influencers. They followed the influencers via multiple platforms, including Facebook (34%), YouTube (52%), Instagram (61%), Twitter (28%), and Snapchat (9%). They followed influencers who specialized in domains like lifestyle (64%), food (45%), fashion (41%), healthy living (40%), and travel (36%). Participants were allowed to choose multiple responses for the last two questions.

Procedure

Upon consenting to participate, participants were asked to answer several screening questions. Those who used social media and who had followed at least one influencer were directed to fill out the rest of the survey questions and were compensated with an additional incentive of $0.10 for answering screen questions and an extra $0.50 for filling out the rest of the survey.

Before the start of the survey, we provided a detailed definition of a social media influencer and asked participants to name the influencer who first came to their minds. The name of that influencer was inserted into the rest of the survey questions. Questions about participants’ social media use, perceptions of the influencer, perceived fairness of interactions with the listed influencer, their interests in products reviewed or promoted by the influencer, as well as their demographic information, were asked. Participants were debriefed in the end.

Measurement

Covariates

Informed by previous studies on parasocial relationships and influencer marketing (e.g., Bernhold and Metzger Citation2020; Bond Citation2018; Lou and Yuan Citation2019), demographical factors—in this case, age and gender—that could affect model testing were included as covariates. In addition, we also considered the duration of social media use as a third covariate, which was found to affect parasocial relationship building in previous studies (e.g., Bond Citation2016). The proposed conceptual model efficiently summarizes our hypotheses (see Figure 1).

Figure 1. Integrated conceptual model.

Display full size

Data Analysis

To test the proposed model, we used SPSS Amos 24 software to conduct structural equation modeling (SEM). SEM allows us to simultaneously test multiple independent variables and dependent variables. We included four dimensions of influencer credibility—expertise, trustworthiness, attractiveness, and similarly—and four dimensions of fairness (i.e., distributive fairness, informational fairness, interpersonal fairness, and procedural fairness) as independent variables, with parasocial relationship being the mediator and interest in products as the dependent variable, while controlling for the effects of age, gender, and time spent on social media. We analyzed both direct effects and indirect effects of the determinants on the dependent variable with bias-corrected bootstrapping (with 95% bias-corrected bootstrap confidence intervals [CIs] based on 2,000 samples) to determine the significance of the indirect effects (Macho and Ledermann Citation2011).

Missing Data Handling

The online questionnaire did not compel the participants to fill in every question. In addition, the participants could choose to drop out. Therefore, there were some missing values in the data set. We carried out a missing pattern analysis. Seven participants had missing values in their answers. We used the option in SPSS to replace missing values with series mean value. Given the small amount of missing values, we believe the missing values caused no or negligible bias in the statistical power of the parameter estimation.

Confirmatory Factor Analysis Measurement Model

The initial measurement model with unit-loading indicators to latent constructs indicated fair model fit, χ² (899) = 2604.37, comparative fit index (CFI) = .89; Tucker–Lewis index (TLI) = .88, standardized root mean square residual (SRMR) = .06, root mean square error of approximation (RMSEA) = .06. Alteration to the model was made by omitting two indicators from the parasocial relationship construct and one indicator from distributive fairness. The two dropped items of parasocial relationship were “I feel sorry for [influencer name] when he/she makes a mistake” and “I find [influencer name] to be attractive.” The item dropped from distributive fairness was “The interactions with [influencer name] about products he/she mentioned are not helpful with my decision-making” (reversed coded). Each had factor loadings lower than .60, the cutoff point (Kline Citation2011). We used the following standards to determine model fit: χ²/df = less than 3 (Kline Citation2011; CFI = greater than .90 (Bentler Citation1992); RMSEA = less than .06 (Hu and Bentler Citation1999). The revised model had acceptable model fit indices, χ² (734) = 1705.21, CFI = .92, TLI = .91, SRMR = .05, RMSEA = .06. All the loadings on the latent constructs were sizable and significant, ranging from .61 to .95, which indicated satisfactory convergent validity (Kline Citation2011). We also conducted a discriminant validity test; average variance extracted (AVE) values for all latent constructs were greater than .50, with the square root of AVE for each construct bigger than its correlation to any other constructs. This indicates adequate discriminant validity among the latent constructs. The correlations of variables in the measurement model are reported in Table 2. Therefore, all the latent constructs had acceptable discriminant validities. Moreover, a collinearity analysis showed no significant levels of collinearity between any sets of the independent variables, with variance inflation factor (VIF) falling between 1 and 5 (Hair et al. Citation2016).

Structural Model Testing

Adopting the previously mentioned model fit indexes, the proposed model showed good model fit with the sample: χ² (852) = 1833.13, χ²/df = 2.15, CFI = .92, TLI = .91, SRMR = .05, RMSEA = .06.

Hypothesis 1 posited that influencer credibility dimensions—(a) expertise, (b) trustworthiness, (c) attractiveness, and (d) similarity—would positively relate to the strength of the parasocial relationship between influencers and their followers. Results showed that expertise (β = .04, p = .52) and trustworthiness (β = .07, p = .39) had no impact on the strength of the parasocial relationship, whereas attractiveness (β = .06, p < .05) and similarity (β = .27, p < .01) positively related to the parasocial relationship strength. Therefore, hypotheses 1(a) and 1(b) were not supported, while hypotheses 1(c) and 1(d) were supported.

Hypotheses 2 posited that fairness—(a) distributive fairness, (b) procedural fairness, (c) interpersonal fairness, and (d) informational fairness—would be positively related to the strength of parasocial relationship. Results revealed that procedural fairness (β = .12, p < .01) and interpersonal fairness (β = .24, p < .01) were positively associated with the strength of a parasocial relationship, whereas informational fairness (β = .10, p = .31) and distributive fairness (β = .18, p = .07) did not affect parasocial relationship. Hypotheses 2(a) and 2(d) were not supported, while hypotheses 2(b) and 2(c) were supported.

Hypothesis 3 postulated about the mediating role of parasocial relationship on the relationship between influencer credibility—(a) expertise, (b) trustworthiness, (c) attractiveness, and (d) similarity—and product interest. Results demonstrated that the parasocial relationship was positively related to product interest (β = .37, p < .05). More importantly, the indirect effects of influencer similarity (β = .10, p < .05) and attractiveness (β = .02, p < .05) on product interest via affecting parasocial relationship were significant, whereas the parasocial relationship mediated neither the effects of expertise (β = .01, p = .55) nor trustworthiness (β = .03, p = .40) on product interest. Despite not being mediated by the parasocial relationship, trustworthiness was found to be negatively related to product interests (β = −.49, p < .001).

Finally, hypothesis 4 focused on the mediating role of parasocial relationship strength on the link between fairness—(a) distributive fairness, (b) procedural fairness, (c) interpersonal fairness, and (d) informational fairness—and product interests. The indirect effects of procedural fairness (β = .04, p < .05) and interpersonal fairness (β = .09, p = .01) on product interests via influencing parasocial relationship were significant. However, the indirect effects of distributive fairness (β = .07, p = .20) and informational fairness (β = .04, p = .33) via parasocial relationship on product interests were not significant. Despite not being mediated by parasocial relationship, distributive fairness was positively related to product interest (β = .52, p < .05). A similar result was also observed with informational fairness (β = .77, p < .001; see Table 3 and Figure 2).

Figure 2. Regression coefficients in the model. Solid line indicates significant path; dotted line indicates nonsignificant path; *p < .05; **p < .01; ***p < .001.

Display full size

Theoretical and Practical Implications

Theoretically, this study considers two critical factors—source credibility and communication justice—in explicating the appeal of influencers among followers. Extant literature on social media influencers or influencer advertising often focuses only on the effects of source characteristics on the effectiveness of sponsored ads posted by influencers or on the formation of follower–influencer relationships. Informed by the fairness conceptualization that originates from organizational communication, this study is the first to propose and identify that followers’ evaluations of the communication process with influencers play an indispensable role in influencer–follower relationships and their consumption-related behaviors. The findings of this study not only add to the literature on source credibility but also extend the repertoire of theoretical framework or conceptualizations that relate to the formation of parasocial relationship in interpersonal communication. In addition, the findings of this study extend our knowledge on the appeal of influencers and the mechanisms through which influencer advertising affect consumer behavior.

Practically, for influencers, it is important to cultivate attractive personae that also signal similarity to followers to strengthen parasocial relationships with their followers. Meanwhile, how followers are being treated by influencers (interpersonal fairness) and whether the interaction with influencers is two-way and somewhat reciprocal (procedure fairness) also matters equally in relationship building. Influencers should showcase adequate etiquette and respect in their interactions with followers and also reciprocate followers’ expectations and emotions to ensure stable relationships with followers. For marketers and brands that are interested in influencer campaigns, besides assessing the reach of influencers (e.g., number of followers) and whether the influencers exhibit due decorum and follow the principle of reciprocity in their interactions with followers, they should also consider the expertise of the influencers in specific domains (distributive fairness) and the ethics that they uphold when sharing information (informational fairness). In particular, only those who deliver accurate and valuable information to followers and those who uphold ethical and moral standards in their sharing are expected to have a positive impact on followers’ interests in the promoted products.

Limitations and Future Research

This study also bears several limitations that point directions for future research. First, to capture participants’ perceptions of an influencer with whom they were familiar, instead of using a specified influencer we asked respondents to name an influencer who first came to their minds. Although we believe this is the most logical way to capture users’ experience in an online survey, and this has been used in prior studies (e.g., Bond Citation2018; Tian and Hoffner Citation2010), we realize that the domain type of the influencers may promise different experiences to followers. Future studies can replicate the study with a more restricted setting—such as influencers in a specific domain or scenario—and find out whether other factors affect the development of parasocial relationships. Second, we recognize that different forms of marketer–influencer collaborations (e.g., product endorsement, product review, gifting, guest blogging) may yield different responses from followers as well, and not all participants in the current study have experienced all kinds of influencer marketing activities. Future studies may consider this difference as a moderator in the model. Third, we did not specify the social media platforms through which participants followed influencers. We acknowledge that different platforms may introduce different features that affect users’ experiences and perceptions of their listed influencers. Future studies can replicate the current study and test the proposed model on one specific media platform. Also, we selected product interests as the outcome that represents marketing objectives in the current study. Future research can examine other marketing- or advertising-related outcomes, such as purchase intentions or actual purchase to validate the role of parasocial relationships. Finally, we used an MTurk sample in the current study, which may not be a representative sample of the general population. We believe future research using more representative samples is needed to test the generalizability of the findings.