How to Make Influencer Advertising Engaging on Instagram: Emotional Storytelling in Sponsored Posts

Abstract

This research investigates how emotional storytelling (ES) in advertising messages in sponsored Instagram posts by microinfluencers compared to macroinfluencers affects social media engagement. In particular, we explore how the tone and intensity of the ES used in sponsored posts affect their engagement. Based on our analysis of 6,122 sponsored posts on Instagram, we find that social media users engage more with sponsored posts with ES than with neutral ones. However, users engage less with sponsored posts with ES when these posts are made by macroinfluencers. Furthermore, results indicate that the effect of tone and intensity of ES on engagement depends on whether sponsored posts are produced by microinfluencers or macroinfluencers. This study provides theoretical insights into how to exploit both the message as well as the source to optimize the effectiveness of influencer advertising. This research also offers practical recommendations to advertisers, agencies, and influencers on how to use ES in sponsored influencer posts to increase social media engagement.

Keywords:

• Agency field data
• emotional storytelling
• influencer advertising
• Instagram
• social media
• social media influencers
• text analytics

Influencer marketing has been growing rapidly, and spending on influencer marketing has almost doubled in the past few years—from $2.42 billion in 2019 to $4.99 billion in 2022 in the United States (Statista 2022). Influencers continue to build a solid base of followers by producing and uploading posts on social media profiles (Ju and Lou 2022; Lou and Yuan 2019). Advertisers increasingly use influencers to help promote their brands and products in sponsored posts on Instagram (Jin, Muqaddam, and Ryu 2019; Lou and Yuan 2019; Yuan and Lou 2020). Influencers integrate advertising messages about brands and products into their sponsored posts and upload them to their social media profiles (De Veirman, Cauberghe, and Hudders 2017; Evans, Phua, Lim, and Jun 2017). In contrast to traditional advertising, influencers produce and upload sponsored posts themselves and control the ultimate message to be communicated (Hughes, Swaminathan, and Brooks 2019). Influencers have been found to be perceived as more intimate and stronger than traditional advertising (Yuan and Lou 2020). In particular, microinfluencers build more intimate and emotionally loaded relationships with their followers (Britt, Hayes, Britt, and Park 2020). Hence, advertisers can encourage but not control which stories influencers tell about their brands and products in sponsored posts.

Existing influencer marketing literature has focused on the size of the influencer’s following (Voorveld 2019), the influencer–follower relationship (Ju and Lou 2022; Marques, Casais, and Anthony Camilleri 2021; Yuan and Lou 2020), and the effect of influencer marketing on advertising efficiency (Britt, Hayes, Britt, and Park 2020; Evans et al. 2017; Lou and Yuan 2019). Moreover, research also focused on the role of advertising messages and appeals in sponsored influencer posts (Gross and Wangenheim 2022; Lou and Yuan 2019). The current study extends previous influencer research and investigates if and how emotional storytelling in sponsored posts affects users’ social media engagement. Building on previous definitions (Luong and Wrzus 2017), we define the term emotional storytelling (ES) as the use of affective experiences in sponsored influencers’ posts. Affective experiences is an umbrella term that includes both the degree of pleasantness and arousal in emotionally loaded experiences with brands and products. The degree of pleasantness refers to the sentiment and presents the emotional tone toward brands and products (Posner, Russell, and Peterson 2005; Russell 1980), such as influencers expressing their pleasant or unpleasant emotionally loaded experiences in sponsored posts. The degree of arousal indicates the strength and magnitude of emotions expressed toward the advertised brands and products (Posner et al. 2005; Russell 1980). For instance, influencers express the strength of the emotionally loaded experience with brands and products in sponsored posts. Examining ES in sponsored posts is essential because emotionally loaded stories provide a convenient mechanism for advertisers to engage users and develop stronger attitudinal and behavioral reactions (Rietveld, van Dolen, Mazloom, and Worring 2020; Wei, Yang, Shoenberger, and Shen 2021).

The foundation of effect of ES in sponsored posts on social media engagement is based on social exchange theory. Social exchange theory describes how behavior between two people is explained by the exchange of resources following a cost-benefit analysis (Blau 1964; Thibaut 1959). The theory suggests that people perceive reciprocity in interpersonal exchange and relationships (Altman 1973). In influencer marketing, social exchange theory captures the influencer–follower relationship (Ju and Lou 2022; D. Y. Kim and Kim 2021). That is, influencers who embed ES in sponsored posts signal authentic recommendations due to the emotionally loaded experience toward brands and products. To express their appreciation and gratitude, users reciprocate by clicking the “Like” button or writing a comment. Hence, influencers who embed ES in sponsored posts will bring more positive social media engagement than those with neutral storytelling in sponsored posts.

In particular, the present study focuses on how ES affects social media engagement with sponsored posts across different types of influencers. Influencers with fewer followers are referred to as microinfluencers while those with a higher number of followers are referred to as macroinfluencers (Voorveld 2019). The positive effect of microinfluencers on influencer marketing has been recorded (Lou and Yuan 2019). That is, microinfluencers foster parasocial relationship and interaction with followers and positively affect social media engagement. In summary, this study examines if ES affects social media engagement by micro- and macroinfluencers. In addition, if ES plays an important role in the efficiency of sponsored posts, the next question is whether the effect on social media engagement by micro- compared to macroinfluencers manifests consistently across the degree of pleasantness and arousal of ES.

In this study, we used a unique data set encompassing 6,122 sponsored Instagram posts by professional influencers. The data were provided by an influencer advertising agency. We find robust and consistent results that social media users engage more with sponsored posts with ES than those with neutral storytelling. However, users engage less with sponsored posts with ES when these are made by macroinfluencers. Furthermore, results indicate that the effect of the degree of pleasantness and arousal of ES on engagement depends on whether sponsored posts are produced by microinfluencers or macroinfluencers.

The findings of this study provide both theoretical and practical contributions. Theoretically, this study advances the extant literature on influencer advertising (De Jans, Cauberghe, and Hudders 2018; De Veirman et al. 2017; Gross and von Wangenheim 2018) by showing how advertisers leverage both the message (i.e., the degree of pleasantness and arousal of ES) as well as the source (i.e., influencers’ number of followers) to influence the effectiveness of sponsored posts. We offer a detailed explanation of the efficiency of the dimensions of ES (i.e., pleasantness and arousal) in sponsored posts, as well as the interplay between ES and influencers’ number of followers. Practically, the findings inform advertisers and agencies about how to strategically use ES for efficient influencer advertising. These findings also offer actionable recommendations for micro- and macroinfluencers in terms of crafting captions to sponsored posts through the tone and intensity of ES.

Theoretical Background

Influencer Advertising on Instagram

Advertisers often launch influencer advertising on multiple platforms simultaneously. Instagram has grown into the most important social media platform for influencer advertising and marketing (“The State of Influencer Marketing 2022: Benchmark Report” 2022). In this fairly new advertising strategy, advertisers strategically incorporate influencers to advertise brands and products on their social media profiles (Tanwar, Chaudhry, and Srivastava 2022). Influencer advertising is an influencer marketing tactic focusing on short-term advertising goals that call consumers to action (Funke 2018). Examples of short-term goals include increasing sales, increasing website traffic, and generating awareness. A commonly used form of influencer advertising is sponsored posts.

Sponsored influencer posts refer to posts with advertising messages produced by influencers and uploaded to their social media profiles but sponsored by advertisers (de Jans et al. 2018; de Veirman et al. 2017). According to Hughes et al. (2019), sponsored influencer posts blur the lines between earned and paid content. Sponsored posts are produced by influencers and uploaded to their social media profiles while advertisers and/or agencies reimburse influencers to do so. Similar to earned content and unlike paid content, influencers have ultimate control of the message to be communicated about brands and products. Hence, influencers blend elements of both earned and paid content when embedding advertising messages in sponsored posts on their social media profiles.

Emotional Storytelling in Sponsored Posts

Sponsored Instagram posts contain text captions with a clear advertising message. Captions provide additional information to the image and often reflect the context and message of images in the form of text. Captions to sponsored posts aim to engage social media users by asking them to become active and react to the sponsored posts. Hence, advertisers use captions as a marketing tool in which they embed their advertising messages about brands and products. Advertisers cannot control the ultimate advertising messages communicated in the caption; however, they can suggest and encourage directions and topics (Hughes et al. 2019).

A common strategy in advertising messages is the use of ES in sponsored posts. In this article, we define ES as the use of affective experiences in sponsored influencers’ posts. We focus specifically on affective experiences in this article because we aim to understand how emotionally loaded experiences affect the success of sponsored posts. We follow the definition by Luong and Wrzus (2017), who use the term affective experiences as an umbrella term that includes both emotions and moods. Emotions, moods, and appraisals involve internal feeling states, psychological as well as behavioral responses, which blurs the lines between the two (Luong and Wrzus 2017; Moors, Ellsworth, Scherer, and Frijda 2013). In the context of advertising, Tellis et al. (2019) found that emotionally loaded content is more likely to be shared on social media. In the context of sponsored posts, influencers tell their stories by expressing their emotionally loaded stories and experiences toward brands and products. For example, they talk about why they liked using a product or which flavors of the product they disliked in advertising messages. We refer to these types of posts as sponsored posts with ES.

ES in sponsored posts holds great promise for advertisers and influencers alike. For advertisers, research has shown that emotions are efficient and persuasive marketing strategies (Holbrook and O’Shaughnessy 1984; Tellis 2003; Tellis et al. 2019), increasing advertising exposure as well as purchase intentions (Poels and Dewitte 2019). Moreover, using emotions in advertising has been found to make consumers feel good, excited, or secure about brands and products (Mizerski and White 1986). Furthermore, a study by Mehta and Purvis (2006) found that emotionally loaded advertising boosts recall. It is unsurprising that advertisers increasingly encourage influencers to tell their emotionally loaded stories and experiences about the advertised brands and products’ sponsored posts.

For influencers, incorporating ES in sponsored posts holds two important benefits. First, ES in sponsored posts adds personal experiences and feelings to the advertising messages. By doing so, influencers build and strengthen their relationships with their followers (Yuan and Lou 2020). Social media users experience a parasocial relationship with influencers and seek their opinion and advice about brands and products (Brewster and Lyu 2020; Marques et al. 2021). As such, social media users believe that they receive honest advice and opinions in sponsored posts by influencers (Boerman and Van Reijmersdal 2020; Childers et al. 2019). Moreover, parasocial interactions between influencers and users have been found to be a key driver for followers’ interest in influencer-advertised brands and products (Yuan and Lou 2020). Therefore, using ES in sponsored posts is indispensable not only for influencers to connect and relate with social media users but also for the efficiency of their sponsored posts.

Second, telling emotionally loaded stories about brands and products in sponsored posts is fundamental to making influencer advertising authentic. For example, Kapitan et al. (2022) found that influencers who demonstrate creative control over their posts are perceived as more believable and authentic, which significantly positively affects their willingness to pay for the advertised brands and products. Moreover, authenticity has been found to be a key driver to establish and maintain the relationship between influencers and their followers (Coco and Eckert 2020). Furthermore, Marwick and Boyd (2011) showed that Twitter posts signal greater authenticity than safely framed public messages. In the context of influencers, sponsored posts are likely to be perceived as more authentic by using emotionally loaded personal experiences about brands and products. In sum, both advertisers and influencers increase the efficiency of their sponsored posts by adding ES to them.

The literature identifies two dimensions of the concept of affective experiences (e.g., see Luong and Wrzus 2017; Russell 2009). We follow the literature and separate ES in sponsored posts into two dimensions. The first one is known as the sentiment in emotional advertising literature (e.g., see Poels and Dewitte 2019). The sentiment reflects which emotions are present in sponsored posts. Moreover, the sentiment also presents the emotional tone and the degree of pleasantness toward brands and products (Posner, Russell, and Peterson 2005; Russell 1980). In the literature, the sentiment is also referred to as valence, the pleasant or unpleasant subjective feeling of an emotionally loaded experience (Shiota, Sauter, and Desmet 2021). In the context of sponsored posts, influencers express degree of pleasantness in advertising messages when telling positive or negative stories. For instance, a sentence like “I love brand X” is a positive expression toward the brand, while the sentence “I dislike brand X” is a negative one; the former is more pleasant than the latter. Influencers navigate the degree of the pleasantness of sponsored posts through the semantic meanings of the words they use to tell their stories. We refer to this dimension as the degree of pleasantness in ES.

The second dimension is known as arousal in the emotional advertising literature (e.g., see Berger and Milkman 2012). The degree of arousal reflects how much emotion is present in the advertising message. Moreover, the degree of arousal reflects the strength and magnitude of emotions expressed toward the advertised brands and products (Posner et al. 2005; Russell 1980). Further, arousal is also referred to the degree of activation experienced through an emotionally loaded experience (Shiota et al. 2021). In sponsored posts, influencers express their degree of arousal about brands and products in the advertising messages when telling stories with more- or less-intensive emotions. Influencers influence the degree of arousal of sponsored posts through the strength and magnitude of the words they use to tell their stories. For instance, when comparing the sentences “I love this conditioner” to “I like this conditioner,” the word love has a higher emotional magnitude than the word like. Hence, the former presents a higher level of arousal compared to the latter. Influencers navigate the degree of arousal in sponsored posts by strength and magnitude of the words they use to tell their stories. We refer to this dimension as the degree of arousal of ES.

In this article, we specifically focus on degree of pleasantness and arousal for two reasons. First, a range of dimensions allows us to capture the effect of the overall degree of emotionality in sponsored posts on engagement. Literature shows that honest and unbiased opinions positively affect users’ responses despite a post being sponsored (Hwang and Jeong 2016). Hence, we use the range of pleasantness and arousal in ES to investigate their effects on engagement. Second, using a spectrum of pleasantness and arousal allows us to get a more diversified picture of emotionally loaded sponsored posts. Our data show that of the emotionally loaded sponsored posts, 7.02% were negative, 78.32% were pleasant, and 98.46% were activating. Thus, we use a definition of ES in sponsored posts that encompasses the influencers’ affective experiences toward brands and products.

Social Media Engagement with Sponsored Posts As Key Performance Indicator in Influencer Advertising

Our key dependent variable for the study is social media engagement. We follow the definition by Swani and Labrecque (2020), who define social media engagement as a mutual process between companies and consumers. Through this mutual beneficial process, companies and consumers co-create brand-related content on social media. Applied to the context of influencer advertising, social media engagement is a mutual process through which advertisers, influencers, and social media users co-create brand-related social media content. Advertisers initiate brand-related content by paying influencers to embed advertising messages into their posts. Influencers then create brand-related content by producing and uploading sponsored posts to their social media profiles. Finally, social media users extend brand-related content by interacting with and contributing to influencers’ sponsored posts. They do so by clicking the “Like” button or writing a comment below a sponsored post. Therefore, advertisers, influencers, and consumers mutually co-create brand-related content and social experiences through sponsored influencer posts on Instagram.

Social media engagement is an important key performance indicator to evaluate the effectiveness of influencer advertising for both advertisers and influencers alike. Advertisers and influencers compare the total number of likes and comments on sponsored posts to those of their competitors to evaluate their performance (Lammenett 2019). More specifically, advertisers assess the total number of likes of and comments on an influencer’s sponsored posts when considering sponsoring their posts. Moreover, influencers rely on the number of likes and comments as advertisers pay influencers based on these interactions (Gross and Wangenheim 2022). Therefore, performance metrics, such as likes and comments, are indispensable in influencer advertising.

In line with this practice, previous research has used likes and comments to operationalize social media engagement (Hughes et al. 2019; Yoon et al. 2018). Social media engagement is particularly relevant for Instagram because the platform has been found to be mainly used for social interaction (Voorveld, van Noort, Muntinga, and Bronner 2018). Social media users use Instagram to pass the time, fill an empty moment, and interact with others (Voorveld et al. 2018). They are looking for interaction and engagement with others on Instagram. Advertisers may consider likes and comments as a return on investment on Instagram; they can also consider the total number of likes of and comments on sponsored posts as consumer engagement with their brands and products. We focus on social media engagement as the interactivity with influencer-sponsored posts and use the number of likes and comments as a proxy for social media engagement.

Linking Emotional Storytelling in Sponsored Posts to Social Media Engagement

Sponsored posts with ES contain emotionally loaded stories about brands and products. These sponsored posts contain personal feelings, experiences, and stories about brands and products. In contrast, sponsored posts with neutral storytelling contain stories that do not tell emotionally loaded experiences or feelings about brands and products in the advertising message, being emotionally neutral toward the advertised brands and products. Marwick and Boyd (2011) showed that Twitter posts that are personal signal greater authenticity than safely framed public messages. Moreover, social media users perceive other users’ and influencers’ posts as useful and authentic with unbiased information and advice (Delbaere, Michael, and Phillips 2021). Further, Kapitan et al. (2022) found that influencers who demonstrate creative control over their posts are perceived as more believable and authentic. Because influencers initiate and engineer their advertising messages, it is likely that social media users perceive sponsored posts with ES as authentic recommendations. In contrast, sponsored posts with neutral storytelling are likely to be perceived as classical advertising.

Social exchange theory provides a theoretical foundation for influencer advertising (D. Y. Kim and Kim 2021). Social exchange theory describes how behavior between two people is explained by the exchange of resources following a cost-benefit analysis (Blau 1964; Thibaut 1959). The theory suggests that when a person invests resources in a relationship, the partner is expected to behave in a way that rewards the person (Kim and Kim 2021). The exchange of the activities in social interactions is reinforced by the two people’s behaviors (Blau 1964; Thibaut 1959). The theory suggests that people perceive a reciprocity in interpersonal exchange and relationships (Altman 1973). In influencer marketing, social exchange theory captures the influencer–follower relationship (Ju and Lou 2022; D. Y. Kim and Kim 2021). Influencers who include personal and emotionally loaded experiences and stories about brands and products in sponsored posts reflect more authentic recommendations than neutral ones. Users, in turn, develop more favorable attitudes and appreciation toward influencer posts with ES. As such, users are likely reciprocating the relationship by liking and commenting on those posts.

Building on these arguments, it is reasonable to assume that social media users engage differently with sponsored posts with ES compared to neutral ones. For example, Hughes et al. (2019) found that emotional and hedonic value positively influences number of comments. In a social media context, authenticity and emotional attachment have been found to be key drivers for following (Kowalczyk and Richards 2016). This arguing is in line with previous work that found developing favorable attitudes toward messages positively affects intentions to click the “Like” button and to write comments (Alhabash, McAlister, Lou, and Hagerstrom 2015). As a result, we hypothesize as follows:

H1: Emotional storytelling in sponsored posts leads to higher social media engagement.

Subsequentially, we focus on the two dimensions of ES. The first dimension, degree of pleasantness, encompasses the tone of emotionally loaded stories about brands and products in sponsored posts. In the context of sponsored posts, influencers express their degree of pleasantness in the advertising messages when telling positive or negative stories. For instance, a sentence like “I love brand X” is a positive expression toward the brand, while the sentence “I dislike brand X” is a negative expression. In the following, we refer to posts using positive emotionally loaded stories about brands and products as pleasant ES in sponsored posts.

The degree of pleasantness in ES varies in sponsored posts. While some ES in sponsored posts is pleasant, other ES is less pleasant. It was shown that advertising likability consists of and depends on pleasantness (Smit, van Meurs, and Neijens 2006). Building on those findings, pleasant ES in sponsored posts might evoke positive, thankful, and appreciative feelings in users. In contrast, unpleasant ES in sponsored posts might evoke negative and unpleasant feelings. Moreover, the reciprocity norm suggests that social exchange occurs by rewards and benefits (Gatignon and Robertson 1986). In the context of influencer marketing, influencers provide a benefit to their users by telling pleasant ES about the advertised brands and products in their sponsored posts. It is reasonable to assume that users want to return the benefit by expressing their feelings of gratitude and appreciation—in form of likes and comments—for pleasant ES in sponsored posts.

In line with this argument, prior research suggests that the degree of pleasantness in ES in sponsored posts might positively affect engagement. For example, Holbrook and Batra (1987) showed that pleasant ads evoke cheerfulness, love, and other pleasant feelings in consumers. In the context of TV ads, pleasantness positively affects consumer hedonism and increases viewing time (Olney, Holbrook, and Batra 1991). Furthermore, Berger and Milkman (2012) showed that positive news is shared more than negative news is shared. Building on these findings from similar contexts, we expect the degree of pleasantness in ES in sponsored posts to positively affect social media engagement. Pleasant ES in sponsored posts is likely to evoke pleasant and affective feelings in consumers which, in turn, encourages them to become active and engage with those posts by clicking the “Like” button or writing a comment. This argument is in line with previous findings in influencer marketing, where the authors suggested that positive emotional responses are likely to be transferred through emotional contagion between influencer and user (Kay, Mulcahy, Sutherland, and Lawley 2022). As a result, we offer a second hypothesis:

H2: The degree of pleasantness in emotional storytelling in sponsored posts positively affects social media engagement.

ES in sponsored posts varies not only by which emotion is present but also how much emotion is present. The second dimension, degree of arousal, encompasses the intensity of emotionally loaded stories about brands and products in sponsored posts. In the context of sponsored posts, influencers express their degree of arousal about brands and products in advertising messages when telling stories that are more emotionally loaded or less emotionally loaded. That is, the strength and magnitude of the words used to tell their stories affects the intensity of ES in sponsored posts. For instance, the sentence “I love this conditioner” has a higher emotionally loaded intensity compared to “I like this conditioner.” In the following, we refer to posts using intensive emotionally loaded stories about brands and products as arousing ES in sponsored posts.

Sponsored posts vary in their degree of arousal in ES. In sponsored posts with high-arousal ES, influencers express their emotionally loaded excitement, intensity, as well as activation toward the advertised brands and products in their stories and experiences. In contrast, influencers tell their stories about the advertised brands and products in a calmer and smoother manner in sponsored posts with low-arousal ES. Belanche et al. (2017) found that high-arousal ads are watched for a longer time than low-arousal ads in the context of online advertising. Because the ES of the advertising message in sponsored posts depends on degree of arousal, users’ reactions to those posts should differ as well.

Drawing on the social exchange theory, we expect the degree of arousal of ES to affect engagement with sponsored posts. Social exchange theory predicts that benefits are reciprocated in relationships (Gatignon and Robertson 1986). If individuals provide benefits to one another, each of them expects to receive equal value in return (Aggarwal 2004). In reciprocal exchange, expressed emotions and affective experiences are argued to have important effects on trust and commitment (Lawler and Thye 1999). Influencers express the strength of felt emotions toward brands and products in sponsored posts with arousing ES, providing higher value to consumers. As such, sponsored posts with arousing ES might be perceived as beneficial in influencer–consumer relationships.

In this vein, high-arousal ES in sponsored posts is expected to be reciprocated by similar feelings of arousal, excitement, or affection. Indeed, arousal in ads has been found to positively affect interestingness and advertising exposure (Olney et al. 1991). In a similar context, Berger and Milkman (2012) found news that evoked more arousing emotions was more likely to be shared than news that evoked fewer arousing emotions. It is likely that high-arousal ES in sponsored posts evokes more arousal and affective user emotions compared to low-arousing ES, which, in turn, positively affects engagement with those posts. We hypothesize:

H3: The degree of arousal in emotional storytelling in sponsored posts positively affects social media engagement.

Emotional Storytelling and Number of Followers

Influencers’ follower counts reflect their degree of public exposure and reach on social media platforms. The number of followers is an indicator of popularity and reflects a social media influencer’s network and group size (de Veirman et al. 2017). Some influencers have fewer followers; others have a higher number. Voorveld (2019) refers to the former as microinfluencers and macroinfluencers to the latter. In contrast to microinfluencers, macroinfluencers are often referred to as famous persons or celebrities (Marques et al. 2021). Microinfluencers have been shown to positively influence parasocial relationships and interactions compared to macroinfluencers (Brewster and Lyu 2020). Hence, social media users perceive a more trustworthy, intimate, and closer relationship to microinfluencers than to macroinfluencers. Indeed, microinfluencers are perceived as more relatable and trustworthy due to their lower follower counts compared to macroinfluencers (Britt, Hayes, Britt, and Park 2020). Moreover, Lou and Yuan (2019) found that influencers’ trustworthiness positively affects followers’ trust in sponsored posts. While sponsored posts by macroinfluencers are perceived as less authentic and more biased, similar posts by microinfluencers are perceived as more authentic and honest (Gross, 2020). One can expect that social media engagement with sponsored posts with ES will vary between micro- and macroinfluencers.

Sponsored posts with ES by macroinfluencers may be less authentic. In practice as well as research, macroinfluencers are often referred to as famous people or celebrities (P2P-Marketing 2019) (Marques et al. 2021). Famous persons’ social media profiles tend to be managed by representatives and are seldom self-maintained (Marwick and Boyd 2011). Similarly, macroinfluencers often have agents and teams in the background who help manage their social media profiles (Lammenett 2019). Similar to traditional celebrity endorsement (McCracken 1989), macroinfluencers might be believed to script emotionally loaded stories in sponsored posts rather than sharing their true emotionally loaded experiences, stories, and feelings about brands and products. Users tend to develop less favorable attitudes toward sponsored posts with ES by macroinfluencers because these posts reflect safely framed recommendations rather than authentic ones. This reasoning is in line with a study by Park et al. (2021), who found that microinfluencers bestow higher perceptions of authenticity on the advertised brands and products compared to their macroinfluencers. As a result, users may engage less with sponsored posts with ES by macroinfluencers than those by microinfluencers. Thus, we expect a higher number of followers to negatively affect social media engagement with sponsored posts with ES. We hypothesize:

H4: A higher number of followers negatively affects social media engagement with sponsored posts with emotional storytelling.

Because users build stronger parasocial relationships with microinfluencers compared to macroinfluencers (Brewster and Lyu 2020), it is likely that they are more involved with sponsored posts by micro- compared to macroinfluencers. Consumer involvement is defined as the levels of interest and personal relevance of a decision and has been identified as antecedents of consumer engagement behaviors (Hollebeek 2011; Hollebeek, Glynn, and Brodie 2014; Mittal 1995; Zaichkowsky 1985). According to information processing theory, consumers put less mental effort into evaluating a message when they are less involved in advertising (Petty, Cacioppo, and Schumann 1983). Consumers evaluate a message less carefully and rely on heuristics, cues, and signals (Petty, Cacioppo, and Heesacker 1981; Tellis 2003); in contrast, strong arguments in a message persuade consumers when they are more involved in advertising (Cacioppo and Petty 1979; Petty et al. 1981). Consistent social media engagement with sponsored posts with pleasant ES as well as arousing ES should vary between micro- and macroinfluencers.

Pleasant ES contains words or sequences of words that direct a message’s semantic meaning. These posts contain positively framed arguments in the form of emotionally loaded experiences, evaluations, advice, or opinions about brands and products. Laurent and Kapferer (1985) stated that when users are involved, they engage in behaviors as active information processing; however, when consumers are not involved, they do not engage in these behaviors. Because users are more involved with microinfluencers, they are likely to perceive pleasant ES as more interesting and relevant compared to macroinfluencers. In this vein, users evaluate the emotional tone in storytelling in sponsored posts by microinfluencers more carefully, which in turn positively affects engagement. Hence, we expect the number of followers to negatively affect social media engagement with sponsored posts with pleasant ES.

H5: A higher number of followers negatively affects social media engagement with sponsored posts with pleasant emotional storytelling.

Arousing ES contains words or sequences of words that direct a message’s semantic intensity. These posts contain arguments in the form of arousing experiences, evaluations, advice, or opinions about brands and products. Arousing ES contains a representation of words that intensify semantic meanings, such as capital letters (e.g., “I DON’T like the brand”), or words with a strong emotional tone (e.g., “I hate the brand!”). When people are less involved with the advertising message, they spend less time on evaluating the content of the message and rely more on cues and heuristics signals (Petty et al. 1981; Tellis 2003). Based on this argument of information processing theory, users are less likely to carefully evaluate advertising messages and are likely to rely on signals when processing sponsored posts by macroinfluencers because users are less involved with them. Hence, signals and cues in ES in sponsored posts may be more persuasive when they are from macroinfluencers compared to microinfluencers. In the context of influencer marketing, semantic intensity is considered a signal or cue because these strongly framed words immediately catch the attention of users. In this vein, users might pay less attention in elaborating the semantic meaning of the advertising message; instead, they might react to the message based on signals and cues in the message, such as semantic intensity, when processing sponsored posts by macroinfluencers. As such, we expect a higher number of followers to positively affect social media engagement with high-arousal ES.

H6: A higher number of followers positively affects social media engagement with sponsored posts with high-arousal emotional storytelling.

Study Overview

Method

To test our hypotheses, we adopted a field data approach. We collected data from an influencer advertising agency that recruits professional influencers for advertisers for sponsored posts on social media. The agency brings together influencers with advertisers for sponsored posts through an interactive selection process. Influencers can apply for sponsored posts by particular advertisers; these advertisers, in turn, can select from the applying influencers. The selected influencers can then finally decide to either accept or not accept the creation of sponsored posts for a particular advertiser. Influencers produce and upload sponsored posts to their profiles, embedding advertising messages into the captions to their posts. The agency focuses on influencers in sports, lifestyle, and health. We analyzed 6,122 sponsored Instagram posts.

In this article, we focus on Instagram for three reasons. First, Instagram is the most important and most impactful social media platform for influencer marketing (Bailis 2019). Instagram has seen increases in its number of monthly active users during recent few years, and those numbers are still increasing (Osman 2019). Second, Instagram is among the world’s leading social media platforms and was among the five most downloaded apps in 2021’s fourth quarter (Chen, 202). Finally, to date, marketing scholars have conducted quantitative research into influencer marketing mostly in the lab (de Jans et al. 2018; de Veirman et al. 2017; Evans et al. 2017), and very little influencer marketing research has been done in the field.

Google Natural Language API

We used a text analytical approach to reliably operationalize our main independent variables. In particular, we used Google Natural Language API (application programming interface) to quantify ES for two reasons. First, we are interested in understanding how influencers exploit the degree of pleasantness and arousal of ES in their captions. The API computes the directions as well as the magnitude of emotions in text on a contiguous scale. Hence, the API allows investigation of emotions in sponsored posts on a more granular level. Second, a major advantage of Google Natural Language API is that the API enables one to analyze text in multiple languages. This feature was indispensable because the data contained sponsored posts in English, German, French, and Italian. Therefore, these algorithms simplify both coding and objectivity.

Sentiment analysis of the Google Natural Language API inspects a given text and identifies emotional opinion and attitude within the text. The sentiment analysis of the API particularly allows to identify if a writer’s attitude is positive, negative, or neutral. The method analyzeSentiment performs the sentiment analysis on given text. This method attempts to determine the overall emotionality in the text and is represented by a numerical sentiment score and a magnitude value. While the sentiment score indicates the overall emotion in a text, the magnitude indicates how much emotion is present in it.

It is important to note that the Google Natural Language API indicates the differences between positive and negative emotions in text but does not identify specific positive and negative emotions. We focus on overall emotionality in sponsored posts. As such, the Google Natural Language API presents a perfect fit because it indicates the overall emotionality of a document with its analyzeSentiment method. In the scope of our study, anger and sadness are both considered negative emotions. Because identifying specific positive or negative emotions was beyond the scope of the study, the feature of identifying specific emotions was not required.

Measurement of Independent Variables

Captions on Instagram include text in the form of words, mentions, hashtags, and emojis. To compute a reliable measurement of sentiment, we excluded emojis from the captions. We also excluded the hashtag character from the caption but left the hashtag’s words in the caption. This is because hashtags are often embedded in the caption and add to a message’s semantic meaning. We excluded all mentions of a caption for the computation of the sentiment because mentions indicate names of other Instagram profiles and do not add to a post’s semantic meaning. To ensure robustness of our results, we estimated additional models using different preprocessing strategies of the caption. Details of our preprocessing strategies can be found in the section on the robustness checks.

We used the automated sentiment analysis of the Google Natural Language API to operationalize the ES of each post. The sentiment score presents the direction of the emotion in the post and indicates if a post contains positive or negative emotions. The sentiment score computed by the API is a continuous variable ranging between −1 (very negative) to +1 (very positive). We use the sentiment score as a proxy to measure the degree of pleasantness in ES in the post.

We also used the automated sentiment analysis to quantify the magnitude of emotion in the post. The magnitude value presents the strength of emotion in the post for both positive and negative emotions. The magnitude value computed by the API is a nonnegative continuous variable. We use the magnitude value as a proxy to measure the degree of arousal of ES in the post.

While the sentiment score indicates which emotion is present in the post, the magnitude value indicates how much emotion is present within the post. Both measurements, sentiment score and the magnitude value, contribute to the overall emotionality present in a caption. For example, a caption with sentiment score and magnitude value close to or equal to 0.0 presents a neutral caption and, as such, indicates sponsored posts with neutral storytelling. To account for the presence of overall emotionality in the caption, we compute a binary variable, emotionality, that equals 1 if the sentiment score or the magnitude value of the captions is not equal to 0.

Model Specification

We estimate the models as follows:

Equation 1: ES on engagement (1) $${\mathrm{Engagement}}_{\mathrm{i}}=\mathrm{constant}+\mathrm{ }{{\mathrm{\alpha }}_1\mathrm{emotionality}}_{\text{i }}+{{\mathrm{\alpha }}_2\mathrm{f}\mathrm{ol}}_{\mathrm{i}}+\mathrm{ }{{\mathrm{\alpha }}_3{\mathrm{fol}}_{\mathrm{i}}\mathrm{emotionality}}_{\mathrm{i}}+{{\mathrm{\alpha }}_4\mathrm{text}}_{\mathrm{i}}+{{\mathrm{\alpha }}_5\mathrm{emoji}}_{\mathrm{i}}+{{\mathrm{\alpha }}_6\mathrm{mention}}_{\mathrm{i}}+{{\mathrm{\alpha }}_7\mathrm{hashtag}}_{\mathrm{i}}+{{\mathrm{\alpha }}_8\mathrm{excl}}_{\mathrm{i}}+{{\mathrm{\alpha }}_9\mathrm{q}\mathrm{ue}}_{\mathrm{i}}+{\mathrm{ }{{\mathrm{\alpha }}_{10}\mathrm{content}}_{\mathrm{i}}+\mathrm{ }{\mathrm{\alpha }}_{11}\mathrm{workhour}}_{\mathrm{i}}+{{\mathrm{\alpha }}_{12}\mathrm{weekend}}_{\mathrm{i}}+\mathrm{ }{\mathrm{\epsilon }}_{\mathrm{i}},$$(1)

Equation 2: Sentiment and magnitude of ES on engagement (2) $${\mathrm{Engagement}}_{\mathrm{i}}=\mathrm{constant}+\mathrm{ }{{\mathrm{\alpha }}_1\mathrm{sentiment}}_{\text{i }}+{{\mathrm{\alpha }}_2\mathrm{magnitude}}_{\mathrm{i}}+{{\mathrm{\alpha }}_3\mathrm{f}\mathrm{ol}}_{\mathrm{i}}+{{{{\mathrm{\alpha }}_4{\mathrm{fol}}_{\mathrm{i}}\mathrm{sentiment}}_{\mathrm{i}}+\mathrm{ }\mathrm{\alpha }}_5{\mathrm{fol}}_{\mathrm{i}}\mathrm{magnitude}}_{\mathrm{i}}+{{\mathrm{\alpha }}_6\mathrm{text}}_{\mathrm{i}}+{{\mathrm{\alpha }}_7\mathrm{emoji}}_{\mathrm{i}}+{{\mathrm{\alpha }}_8\mathrm{mention}}_{\mathrm{i}}+{{\mathrm{\alpha }}_9\mathrm{hashtag}}_{\mathrm{i}}+{{\mathrm{\alpha }}_{10}\mathrm{excl}}_{\mathrm{i}}+{{\mathrm{\alpha }}_{11}\mathrm{q}\mathrm{ue}}_{\mathrm{i}}+{\mathrm{ }{{\mathrm{\alpha }}_{12}\mathrm{content}}_{\mathrm{i}}+\mathrm{ }{\mathrm{\alpha }}_{13}\mathrm{workhour}}_{\mathrm{i}}+{{\mathrm{\alpha }}_{14}\mathrm{weekend}}_{\mathrm{i}}+\mathrm{ }{\mathrm{\epsilon }}_{\mathrm{i}},$$(2) where i indexes the post that has been uploaded to an Instagram profile (i = 1, 2, 3, . . . 6,122). In Equation 1, we investigate the effect of the presence of emotionality in sponsored posts on engagement. More specifically, we compare how users engage with sponsored posts with ES compared to neutral stories. In Equation 2, we look at emotionality in sponsored posts at a more granular level. In particular, we explore the effect of the sentiment and magnitude of ES in sponsored posts on their engagement.

The dependent variable, engagement, is the sum of the number of likes and comments a post has. On the right-hand side of the equation, the model consists of the main independent variables. We operationalized emotionality in sponsored posts (emotionality) as a dummy variable that equals 1 if the post contains any degree of (un)pleasantness or arousal of ES in the advertising message and 0 if the post is emotionally neutral. In Equation 2, we quantified the degree of pleasantness in ES, sentiment, as a continuous variable. The variable sentiment ranges between −0.9 = Very negative and +0.9 = Very positive in our data set. We quantified the degree of arousal of ES, magnitude, as a continuous nonnegative variable. The variable magnitude ranges from 0 to 16 in our data set. We included a continuous variable, fol, which indicates the number of followers of the influencer who uploaded the sponsored post to his or her profile. We included the interaction terms between influencer’s follower count and emotional variables (fol × emotionality, fol × sentiment, fol × magnitude) to explore whether the effect of ES in sponsored posts on engagement varies between micro- and macroinfluencers.

In an alternative model, we also included the interaction term between the sentiment and magnitude of ES (sentiment × magnitude) as well as interaction with the number of followers (fol × sentiment × magnitude). We included the interaction term because ES varies in sentiment and magnitude at the same time. For instance, emotional experiences and opinions may be pleasant and arousing. The expression “I love this shampoo” has a sentiment of 0.9 (very positive) and a magnitude of 0.9. However, we found no significant effects when including the interaction term. Moreover, the main variables’ results did not change in direction and significance when we included these interaction terms. Hence, we excluded these two variables from further analysis.

We added some control variables into our model to remove extraneous influence from the dependent variables. First, social media users perform less online advertising searching during weekends than weekdays (De Vries, Gensler, and Leeflang, 2012; Rutz and Bucklin, 2011). We included a weekend dummy (weekend) and a working-hours dummy (workhour) as covariates in the model because it may be that users visit influencers’ profiles or see their posts more during leisure time than during work time, which may affect higher engagement with a post. Second, a post’s content type positively affects its popularity (de Vries et al., 2012). We included a dummy that indicated whether a post has static or dynamic content (content). We also controlled for several post characteristics, such as text length (text), number of emojis (emoji), number of mentions (mention), number of hashtags (hashtag), number of exclamations (excl), and number of question marks (que). Table 1 contains the correlation of all variables.

Table 1. Correlation matrix among the variables.

Variables	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16
1 Likes	1
2 Comments	0.47	1
3 Engagement	1	0.48	1
4 Emotionality	0.02	0.01	0.02	1
5 Magnitude	0.00	0.02	0.00	0.15	1
6 Sentiment	0.04	−0.02	0.04	0.42	0.15	1
7 Fol	0.74	0.43	0.74	0.04	0.05	0.06	1
8 Text	0.00	−0.02	0.00	0.05	0.81	0.02	0.06	1
9 Emoji	−0.03	0.03	−0.03	0.01	0.28	0.01	−0.02	0.33	1
10 Mention	0.00	−0.06	0.00	−0.01	0.15	0.01	−0.01	0.25	0.15	1
11 Hashtag	−0.02	−0.12	−0.02	0.00	0.19	0.03	−0.01	0.43	0.03	0.25	1
12 Excl	0.00	0.01	0.00	0.07	0.41	0.05	0.01	0.37	0.18	0.12	0.04	1
13 Que	−0.01	0.07	−0.01	−0.06	0.35	−0.12	0.03	0.33	0.18	−0.03	0.03	0.12	1
14 Content	−0.02	−0.08	−0.02	0.01	−0.01	0.01	−0.01	0.00	−0.03	0.01	0.00	0.03	−0.01	1
15 Workhour	−0.03	0.01	−0.03	0.03	0.08	0.03	0.00	0.08	0.05	−0.06	0.01	0.03	0.04	−0.01	1
16 Weekend	0.00	−0.02	0.00	0.00	0.00	0.00	0.02	0.00	−0.01	0.00	0.03	0.01	0.00	0.00	0.04	1

Estimation Strategy

We first estimated the effect of emotions in sponsored posts on engagement with and without control variables (Equation 1). We then estimated the effect of the sentiment and magnitude of ES in sponsored posts on engagement with and without control variables (Equation 2). Table 2 summarizes the estimation results using a negative binomial regression. In our data, the variance of our dependent variable engagement (VAR = 12,465,261) exceeded its mean (M = 1,658.86), indicating overdispersion. An overdispersion test confirmed overdispersion in our data (α = 1,211, p value < 0.001). Thus, we used a negative binomial regression to account for the overdispersed data (Greene 2020). Alternative estimations for count data are Poisson regression and zero-inflated regression models (Greene 2020). Results remained unchanged using a Poisson regression. The Poisson regression’s results are presented as a robustness check in Table 3. We excluded the zero-inflated regression model from our analysis because all posts in the data had at least one like or comment.

Table 2. Estimation results: Negative binomial regression.

Variables	Overall Emotionality		Sentiment and Magnitude
	Without CVs	With CVs	Without CVs	With CVs
Emotionality	0.149***	0.162***
	(0.037)	(0.036)
Fol × Emotionality	−0.037***	−0.033***
	(0.005)	(0.005)
Sentiment			0.247***	0.178***
			(0.048)	(0.049)
Fol × Sentiment			−0.045***	−0.040***
			(0.006)	(0.006)
Magnitude			−0.069***	0.021
			(0.011)	(0.015)
Fol × Magnitude			0.003**	0.001
			(0.001)	(0.001)
Control variables
Fol	0.165***	0.163***	0.143***	0.145***
	(0.005)	(0.005)	(0.003)	(0.003)
Text		−0.146***		−0.183***
		(0.020)		(0.029)
Emoji		0.057***		0.055***
		(0.015)		(0.014)
Mention		−0.005		−0.001
		(0.012)		(0.012)
Hashtag		−0.035*		−0.028^+
		(0.014)		(0.016)
Excl		−0.012		−0.018
		(0.013)		(0.013)
Que		0.016		0.012
		(0.011)		(0.012)
Content		−0.525***		−0.508***
		(0.098)		(0.097)
Workhour		0.026		0.022
		(0.023)		(0.024)
Weekend		−0.021		−0.019
		(0.028)		(0.028)
Constant	6.216***	6.191***	6.391***	6.240***
	(0.034)	(0.035)	(0.025)	(0.033)
Observations	6,122	6,122	6,122	6,122
AIC	97,054	96,832	96,985	96,828

Note. Fol are reported in 10,000 followers. We reported standardized coefficients and robust standard errors. CVs: control variables; AIC: Akaike information criterion.

⁺p < 0.1; *p < 0.05; **p < 0.01; ***p < 0.001.

Table 3. Estimation results: Linear regression and the Poisson regression.

Variables	Linear Regression	Poisson Regression
Sentiment	0.144*	0.155*
	(0.061)	(0.062)
Fol × Sentiment	−0.038***	−0.020***
	(0.007)	(0.005)
Magnitude	0.019	−0.062**
	(0.017)	(0.022)
Fol × Magnitude	0.004**	0.010***
	(0.001)	(0.001)
Control variables
Fol	0.139***	0.074***
	(0.004)	(0.003)
Text	−0.186***	−0.190***
	(0.029)	(0.045)
Emoji	0.079***	0.047*
	(0.015)	(0.019)
Mention	0.056***	−0.021
	(0.015)	(0.020)
Hashtag	−0.029+	0.026
	(0.017)	(0.022)
Excl	−0.035*	0.010
	(0.015)	(0.021)
Que	0.039**	0.018
	(0.015)	(0.019)
Content	−0.878***	−0.347+
	(0.085)	(0.179)
Workhour	0.047+	−0.008
	(0.028)	(0.039)
Weekend	−0.054	−0.053
	(0.033)	(0.047)
Constant	5.829***	6.770***
	(0.039)	(0.049)
Observations	6,122	6,122
R^2	0.428	—
Adjusted R^2	0.427	—
AIC	—	6,273,343

Note. Fol are reported in 10,000 followers. We reported standardized coefficients and robust standard errors. AIC: Akaike information criterion.

⁺p < 0.1; *p < 0.05; **p < 0.01; ***p < 0.001.

Descriptive Results

We analyzed 6,122 sponsored Instagram posts; 43% were uploaded during working hours; 23% during weekends; 81% contained emojis in their captions; and 75% contained at least one hashtag. All the posts contained additional text characters such as question marks, exclamations, or regular words. The average number of total characters per caption was 406.

The mean number of likes per post was 1,625.47 (SD = 3,514.34, median [Mdn] = 686); the number of likes ranged between 1 and 69,585. The mean number of comments per post was 33.39 (SD = 34.14, Mdn = 20). The number of comments ranges from 0 to 150. The mean engagement per post was 1,658.86 (SD = 3,530.62, Mdn = 720); engagement ranged from 1 to 69,722. The mean number of followers was 44,890.25 per post (SD = 68,647.15, Mdn = 18,386). The number of followers per Instagram profile ranged from 1,016 to 505,237.

Of the posts, 85.34% were emotional and 14.66% were neutral. Of the emotional posts, 7.02% were negative; 78.32% were pleasant; and 98.46% were activating. The mean engagement was 1,695.33 (SD = 3,664.44, Mdn = 739) for emotional posts, compared to 1,446.43 (SD = 2,610.10, Mdn = 598) for nonemotional posts.

Estimation Results

First, our results supported hypothesis 1: ES in sponsored posts led to higher social media engagement. The coefficient of the variable emotionality was positive and highly significant (β = 0.162, p < 0.001). This indicates that social media users were more likely to like and comment on sponsored posts with ES compared to those with neutral storytelling.

Second, the coefficient of the sentiment of ES (sentiment) was positive and highly significant (β = 0.178, p < 0.001), showing that the direction of ES related positively to engagement. Hence, social media users engaged more with sponsored posts with more pleasant ES than less pleasant ES. Hypothesis 2—the degree of pleasantness in ES in sponsored post positively affects engagement—was supported.

Hypothesis 3 posited that the degree of arousal of ES in sponsored posts would positively affect engagement. As expected, the coefficient of the magnitude of ES (magnitude) was positive (β = 0.021). In contrast to what we expected, the magnitude of ES did not significantly affect engagement (p > 0.1). Our findings did not support hypothesis 3.

Regarding number of followers, we observed a negative relationship between number of followers and sponsored posts with ES (fol × emotionality). The coefficient of the interaction term was negative and highly significant (β = −0.033, p < 0.001). The findings support hypothesis 4—that a higher number of followers negatively affects social media engagement with sponsored posts with ES.

Further, the interaction coefficient between number of followers and sentiment of ES was negative and highly significant (β = −0.040, p < 0.001). Indeed, a higher number of followers negatively affected engagement with sponsored posts with pleasant ES, which supported hypothesis 5.

However, our results do not support hypothesis 6—that a higher number of followers positively affects social media engagement with sponsored posts with arousing ES. The interaction coefficient between a higher number of followers and the magnitude of ES was positive and insignificant (β = 0.001, p > 0.1). Because all control variables were mainly used for capturing and controlling factors that may influence our dependent variable, we conduct no further interpretation here.

Robustness Checks

To increase confidence in our results, we ran several robustness checks. First, we used an alternative regression model. Marketing research with a similar data structure has used log-transformed dependent variables in linear regression models (de Vries et al. 2012; Sabate, Berbegal-Mirabent, Cañabate, and Lebherz 2014). Following Sabate et al. (2014), we ran a linear regression model with a log-transformed dependent variable. The distribution of the dependent variable, engagement, was skewed toward the right, with a skewness of 7.93. A log-transformation reduced skewness to the left. The estimation results showed further support for our findings. The results are presented in Table 3. In sum, the results of the robustness checks showed support for our main model’s findings.

Second, we used alternative dependent variables for engagement. First, we looked at the number of likes and comments separately. Sponsored posts with pleasant ES received significantly more likes (β = 0.190, p < 0.001) but marginally significantly fewer comments (β = −0.101, p < 0.1). The difference may be that likes and comments represent different levels of digital and social media engagement. While shares and likes are often referred to as shallow engagement, comments are referred to as deep engagement (Yoon et al. 2018). Berger and Milkman (2012) argued that most people prefer to be known as someone who shares upbeat stories or makes others feel good. The same may be true for clicking the “Like” button or writing a comment. More users prefer to click the “Like” button on sponsored posts with pleasant ES because this reflects positively on the liker. In line with this, users prefer writing comments to express feelings of support and understanding when seeing sponsored posts with less pleasant ES, because this reflects positively on the commenter. Confirming our main results, a higher number of followers negatively affected the number of likes of and comments on sponsored posts with pleasant ES. In line with this, the result for more arousing ES in sponsored posts remained the same in direction and significance for both the number of likes and comments. Second, we computed an alternative engagement variable. The research argues that comments are worth more than likes. For instance, Kim and Yang (2017) stated that a comment is worth as much as seven likes; this is in line with practice. Because advertisers pay up to five times more to influencers for a comment than for a like, we transformed the number of comments in number of likes by multiplying the number of comments by five. We computed an alternative engagement variable as the sum of the number of likes and the number of comments: however, the number of comments were expressed in number of likes. The coefficients show further support of the estimation results in the main model. Table 4 presents the estimation results of the negative binomial regression with different dependent variables.

Table 4. Estimation results: Negative binomial regression with alternative DVs.

Variables	Likes As DV	Comments As DV	Alternative Engagement As DV
Sentiment	0.190***	−0.101+	0.140**
	(0.050)	(0.056)	(0.047)
Fol × Sentiment	−0.042***	−0.014**	−0.036***
	(0.006)	(0.005)	(0.006)
Magnitude	0.021	0.021	0.021
	(0.015)	(0.014)	(0.014)
Fol × Magnitude	0.001	0.001	0.001
	(0.001)	(0.001)	(0.001)
Control variables
Fol	0.148***	0.061***	0.138***
	(0.003)	(0.002)	(0.003)
Text	−0.187***	−0.086***	−0.170***
	(0.030)	(0.023)	(0.026)
Emoji	0.055***	0.049***	0.055***
	(0.014)	(0.012)	(0.013)
Mention	0.00002	−0.039**	−0.005
	(0.012)	(0.014)	(0.011)
Hashtag	−0.025	−0.106***	−0.037*
	(0.016)	(0.016)	(0.015)
Excl	−0.020	0.025+	−0.012
	(0.014)	(0.013)	(0.013)
Que	0.010	0.058***	0.017
	(0.012)	(0.012)	(0.011)
Content	−0.507***	−0.610***	−0.515***
	(0.098)	(0.107)	(0.092)
Workhour	0.020	0.062*	0.025
	(0.024)	(0.027)	(0.022)
Weekend	−0.017	−0.082**	−0.027
	(0.028)	(0.032)	(0.027)
Constant	6.195***	3.149***	6.402***
	(0.034)	(0.035)	(0.031)
Observations	6,122	6,122	6,122
AIC	96,463	54,297	98,218

Note. Fol are reported in 10,000 followers. We reported standardized coefficients and robust standard errors. DV: dependent variable; AIC: Akaike information criterion.

⁺p < 0.1; *p < 0.05; **p < 0.01; ***p < 0.001.

Third, the sentiment and the magnitude of ES might vary based on the different characteristics of a sponsored post. We estimated results using a negative binomial regression with alternative preprocessing methods of a post. First, we computed the sentiment and the magnitude of ES by replacing emojis with their corresponding text descriptions. To do so, we used an encoding library for emojis that contains descriptions of emojis. Second, we computed the sentiment and the magnitude of ES, describing emojis in words and retaining all hashtag characters. Finally, we computed the sentiment and the magnitude of ES using the original sponsored post. Table 5 summarizes the estimation results using a negative binomial regression with alternative preprocessing methods of a post. The results show that coefficients were the same in direction as in the main model for all preprocessing methods. Significance of the coefficients was the same as in the main model except for the significance of the coefficient of sentiment.

Table 5. Estimation results: Negative binomial regression with different post processing.

Variables	Emojis in Words, No Tags	Emojis in Words, with Tags	Original Post
Sentiment	0.158**	0.089+	0.135**
	(0.050)	(0.047)	(0.046)
Fol × Sentiment	−0.044***	−0.031***	−0.032***
	(0.006)	(0.006)	(0.006)
Magnitude	0.021	0.017	−0.003
	(0.014)	(0.014)	(0.014)
Fol × Magnitude	0.001	0.001	0.001
	(0.001)	(0.001)	(0.001)
Control variables
Fol	0.147***	0.142***	0.143***
	(0.003)	(0.003)	(0.003)
Text	−0.182***	−0.177***	−0.151***
	(0.029)	(0.029)	(0.028)
Emoji	0.057***	0.057***	0.057***
	(0.015)	(0.015)	(0.015)
Mention	−0.002	−0.001	−0.003
	(0.012)	(0.012)	(0.012)
Hashtag	−0.027+	−0.028+	−0.034*
	(0.016)	(0.016)	(0.016)
Excl	−0.019	−0.018	−0.014
	(0.013)	(0.014)	(0.014)
Que	0.009	0.010	0.014
	(0.012)	(0.012)	(0.012)
Content	−0.508***	−0.510***	−0.509***
	(0.096)	(0.096)	(0.096)
Workhour	0.022	0.023	0.026
	(0.024)	(0.024)	(0.024)
Weekend	−0.019	−0.019	−0.021
	(0.028)	(0.028)	(0.028)
Constant	6.242***	6.273***	6.293***
	(0.032)	(0.031)	(0.032)
Observations	6,122	6,122	6,122
AIC	96,825	96,839	96,841

Note. Fol are reported in 10,000 followers. We reported standardized coefficients and robust standard errors. AIC: Akaike information criterion.

⁺p < 0.1; *p < 0.05; **p < 0.01; ***p < 0.001.

Next, we estimated our results with different subsamples and control variables. First, we used a subsample that excluded dynamic content, because social media users may engage differently with dynamic compared to static content. It may be that social media users engage passively with dynamic content because they see or watch it (e.g., similar to watching TV) rather than actively engaging with it. We excluded 2.73% of the posts that were of dynamic content. Second, we used a subsample that excluded posts uploaded during working hours. Social media users may spend more time on Instagram during leisure time than during working hours, which may affect engagement. We excluded 32.01% of the posts that were uploaded during working hours. In both subsamples, we used the total number of characters (nchar) as the control variable. In each subsample, we controlled for content and workhour/weekend, respectively. The results for the sentiment of ES were the same in both direction and significance, including the interaction term. However, the coefficient of the interaction term of the magnitude of ES and followers changed slightly in significance. Table 6 presents the estimation results using the negative binomial regression.

Table 6. Estimation results: Negative binomial regression with different subsamples.

Variables	Posts without Dynamic Content	Posts without Working Hours
Sentiment	0.178***	0.284***
	(0.049)	(0.059)
Fol × Sentiment	−0.044***	−0.049***
	(0.006)	(0.007)
Magnitude	0.029*	−0.010
	(0.015)	(0.015)
Fol × Magnitude	0.002*	0.002+
	(0.001)	(0.001)
Control variables
Fol	0.144***	0.144***
	(0.003)	(0.003)
Nchar	−0.193***	−0.138***
	(0.021)	(0.024)
Content	—	−0.545***
	—	(0.127)
Workhour	0.023	—
	(0.023)	—
Weekend	−0.028	—
	(0.028)	—
Constant	6.235***	6.283***
	(0.032)	(0.032)
Observations	5,955	4,162
AIC	94,356	65,841

Note. Fol are reported in 10,000 followers. We reported standardized coefficients and robust standard errors. AIC: Akaike information criterion.

⁺p < 0.1; *p < 0.05; **p < 0.01; ***p < 0.001.

Results of all the previously mentioned robustness checks showed further support for our findings of the main model.

Discussion

Advertising with influencers on social media has continuously grown during the past decade. The current study examined if and how ES affects the effectiveness of influencer advertising. More specifically, we explored how users engage with sponsored posts with ES compared to those with neutral storytelling. To draw a more comprehensive picture on ES in sponsored posts, we investigated how the different dimensions of ES—the degree of pleasantness and arousal—influence social media engagement with sponsored posts. We further explained how these effects vary between micro- and macroinfluencers. The findings of this research advance the effectiveness of influencer marketing (Hughes et al. 2019) as well as the literature on advertising messages in sponsored posts (Gross and Wangenheim 2022; Lou and Yuan 2019; Tafesse and Wood 2021) and the role of follower numbers of influencers in advertising (de Veirman et al. 2017; Voorveld 2019; Voorveld et al. 2018).

The first major finding shows how ES in sponsored posts affects social media engagement. This finding adds to literature on the role of influencer as an advertising strategy in sponsored posts (e.g., see Schouten, Janssen, and Verspaget 2020). Our results show that users engage more with sponsored influencer posts with ES compared to neutral posts. Potential explanations can be found in social exchange theory (Cook, Cheshire, Rice, and Nakagawa 2013). Influencers who include personal and emotionally loaded experiences and stories about brands and products in sponsored posts reflect more authentic recommendations than neutral ones. This may imply that users, in turn, develop more favorable attitudes and appreciation toward influencer posts with ES. As such, users are likely reciprocating the relationship by liking and commenting on those posts. This finding is in line with previous research that shows the relationship between influencers and followers is indeed reciprocal based on a cost-benefit analysis (Ju and Lou 2022).

Another important finding of the study relates to the dimensions of ES. We find that the degree of pleasantness—but not the degree of arousal—positively affects social media engagement with sponsored posts. We do not find empirical evidence that the degree of arousal of ES in sponsored posts affects social media engagement. We find no support that users’ engagement varies with the intensity of ES in sponsored posts. One potential explanation for this finding might be that users give first or even more attention to the tone in ES toward the advertised brand or product rather than the strength and magnitude of expressed emotions in storytelling. When being exposed to sponsored posts, users might first evaluate the sentiment and emotional tone (i.e., the degree of pleasantness), which then influences how they evaluate advertising messages in sponsored posts. This is in line with our findings that show users engage more with more-pleasant ES compared to unpleasant ES in sponsored posts.

Another important finding concerns the important role of number of followers on the effects of ES in sponsored posts on social media engagement. We show that number of followers negatively affects social media engagement with sponsored posts with ES, which is in line with the findings of a study by Park et al. (2021). Sponsored posts with ES by macroinfluencers may be less authentic. Macroinfluencers often have agents and teams in the background who help manage their social media profiles (Lammenett 2019). Similar to traditional celebrity endorsement (McCracken 1989), macroinfluencers might be believed to script their emotionally loaded stories in sponsored posts rather than sharing their true emotionally loaded experiences, stories, and feelings about brands and products. It is not surprising that users tend to develop less favorable attitudes toward sponsored posts with ES by macroinfluencers because these posts reflect safely framed recommendations rather than authentic ones.

Moreover, our findings show that number of followers negatively affects social media engagement with sponsored posts with pleasant ES. Because users are more involved with microinfluencers, they are likely to perceive their sponsored posts with pleasant ES as more interesting and relevant compared to macroinfluencers. Consumers put less mental effort into evaluating a message when they are less involved in advertising (Petty et al. 1983), evaluating a message less carefully and rely on heuristics, cues, and signals (Petty et al. 1981; Tellis 2003); in contrast, strong arguments in a message persuade consumers when they are more involved in advertising (Cacioppo and Petty 1979; Petty et al. 1981). According to information processing theory, users might elaborate the emotional tone in storytelling in sponsored posts by microinfluencers more carefully which, in turn, positively affects engagement.

Finally, the current research explores factors that determine the effectiveness of influencer advertising. Previous research in this field focused on understanding the influencer–follower relationship (Ju and Lou 2022; Marques et al. 2021; Yuan and Lou 2020) or the effects of influencers on attitudes and intentions toward brands and products (de Jans et al. 2018; Evans, Hoy, and Childers 2018). In contrast, we add to the knowledge if and how ES in sponsored influencer posts affects social media engagement. We have refined and extended previous literature on influencer advertising efficiency (Britt et al. 2020; Evans et al. 2017; Lou and Yuan 2019; Park et al. 2021; Yuan and Lou 2020), as well as responded to the call for more research on influencer marketing (Voorveld 2019).

Theoretical Implications

This study yields rich theoretical contributions. First, this study contributes to current literature on influencer advertising. Previous research in influencer advertising investigated one dimension of the effectiveness of sponsored posts (Evans et al. 2017; Lou and Yuan 2019; Thorson and Rodgers 2006; Yuan and Lou 2020). In the present study, we explored how both advertisers and influencers might leverage both dimensions—the message (i.e., ES) and the source (i.e., type of influencer)—to affect social media engagement with sponsored posts. Previous research focused on credibility when investigating the influencer as a source (Lou and Yuan 2019; Yuan and Lou 2020); we add to this research by investigating another important characteristic of influencers: number of followers. While previous research investigated emotions in advertising (Barger, Peltier, and Schultz 2016; Holbrook and Batra 1987; Olney, Holbrook, and Batra 1991) or the effects of micro- and macroinfluencers on advertising (Britt, Hayes, Britt, and Park 2020; Evans et al. 2017; Lou and Yuan 2019; Yuan and Lou 2020), little research exists combining both. We add to the body of literature by exploring message-specific and source-specific characteristics makes influencer advertising efficient. This is in line with previous work reflecting the importance of both message and source when using sponsored influencer posts in advertising (Gross and Wangenheim 2022; Tanwar et al. 2022).

Second, previous studies have focused on influencer advertising outcomes from a single perspective, such as brand awareness, purchase intentions, or parasocial interaction (Lou and Yuan 2019; Marques et al. 2021; Yuan and Lou 2020). Our study adds to the influencer literature with unique insights using an outcome that affects both advertisers and influencers at the same time. While likes and comments are important indicators that reflect an influencer’s attractiveness as a sponsorship partner for companies, likes and comments also represent the financial income of influencers. Exploring the effects of ES on the number of likes and comments affect advertisers and influencers alike. We show how both advertisers and influencers can exploit ES in sponsored posts to increase their effectiveness.

In a broader sense, we extend the literature on social exchange theory (Cook et al. 2013). We find that ES in sponsored posts positively affects social media engagement. This result suggests that users engage more with sponsored posts with emotionally loaded stories and experiences compared to neutral ones. The reason for this finding can be found in social exchange theory. In line with predictions from social exchange theory, we argue that users reciprocate the relationship by clicking the “Like” button or writing a comment. Influencers who include personal and emotionally loaded experiences and stories about brands and products in sponsored posts provide benefits to users in the form of authentic recommendations. To reciprocate the benefits, users engage more with sponsored posts with ES to express their appreciation for authentic recommendations.

Finally, this study responds to the call for more research investigating the role of emotions in advertising (e.g., see Poels and Dewitte 2019) as well as influencer advertising and marketing (e.g., see Voorveld 2019). We provide two important contributions that help build influencer marketing as a unique literature stream by exploring its efficiency from a dual perspective (Hughes et al. 2019; Oh, Keller, and Neslin 2020; Voorveld 2019). First, our findings show that advertisers should adapt the strategic use of ES depending on the type of influencer (microinfluencers versus macroinfluencers) those posts the advertisers are sponsoring. Second, influencers should strategically use ES when embedding advertising messages into their posts. Our research suggests that goals of both advertisers and influencers should be considered when investigating the efficiency of influencer advertising. Hence, future research should explore the effectiveness of other dimensions in influencer advertising from a dual perspective.

Practical Implications

Our study identifies the important influence of ES in understanding engagement behavior with sponsored posts for advertisers and influencers alike. From an advertiser perspective, users like and comment more on sponsored posts with ES compared to those without ES. Our findings reveal that users like and comment more on sponsored posts with a higher degree of pleasantness in ES; however, we could not find any significant effects on liking and commenting behavior on sponsored posts with a higher degree of arousal of ES. As such, influencers should focus on working with the tone rather than intensity when using ES in sponsored posts. Advertisers are advised to encourage influencers to tell their personal experiences and attitudes about their brands and products in sponsored posts. For example, influencers could specifically describe their personal relation and experience when using the advertised brand or product.

From an agency perspective, our findings suggest that agencies should advise different types of influencers based on the goal of advertising messages the advertisers encourage to see. Agencies are advised to recommend microinfluencers to advertisers when those are looking to work with ES in sponsored influencer posts. Agencies should cooperate with microinfluencers when advertisers are looking for pleasant ES about their brands and products in sponsored influencer posts. In contrast, agencies should rely on macroinfluencers when advertisers are looking for more neutral ES in sponsored influencer posts. By doing so, they may optimize the goals of both advertisers and influencers.

From an influencer perspective, our study shows that the number of followers affects engagement behavior with sponsored posts. Our findings reveal that users like and comment more on sponsored posts with ES when being posted by microinfluencers. Moreover, users like and comment more on sponsored posts with pleasant ES by micro- compared to macroinfluencers. As a consequence, influencers should focus more on tone rather than strength when telling their emotionally loaded stories about brands and products in sponsored posts. While pleasant ES in sponsored posts results in more social media engagement, unpleasant ES in sponsored posts may positively affect followers’ estimation of influencers’ realness and credibility. Hence, influencers are advised to use both pleasant and unpleasant emotionally loaded stories, even though doing so may result in lower social media engagement.

Limitations and Future Research

The current study has limitations. First, this study relied on metrics from Google Natural Language API to detect the degree of pleasantness and arousal of ES in sponsored posts. However, there might be different ways to measure ES in sponsored posts. Other tools (e.g., LIWC, VADER lexicon; Hutto and Gilbert 2014) may rely on different language lexicons, which may yield other results. Future research could then replicate our results by using different metrics and tools to measure ES and its dimensions. In addition, future work could further verify the external validity of our results by doing so.

Second, the primary goal of the current study was to explore how ES could be leveraged in sponsored posts from a dual perspective. We focused on social media engagement as the main key performance indicator for both influencers and advertisers (Lammenett 2019). Yet other types of outcomes might be valuable to both advertisers and influencers. For example, future research could explore how followers talk about the advertised brands and products in the comments below the posts, as well as how to mediate those conversations. Future studies might consider the sentiment and magnitude of ES in the user comments as a potential alternative outcome variable.

Third, the present study investigated if and how ES in sponsored posts affects social media engagement. However, ES might be investigated on a more granular level. The effects of ES in sponsored posts on social media engagement might differ for different types of emotions. For example, it is likely that users react more actively to sad sponsored posts to show sympathy, which, in turn, might positively affect social media engagement. Future research might also consider comparing specific or discrete emotions in storytelling and their effects on social media engagement with sponsored posts.

Further, to achieve this breadth of data, we had to accept a few limitations in our data, which open fruitful avenues for future research. Effect sizes across the analysis are small to medium, suggesting that other factors should be considered in the future with more comprehensive models aimed to produce larger effect sizes. Moreover, a small risk of unobserved heterogeneity remains despite many efforts of controlling for potential confounders. Of the posts, 43.45% had a character count of 255. Some of those posts were cut off at character 255, and some had exactly a number of characters of 255. Cutoffs were due to some technical issues at Instagram’s API. Excluding all posts with a number of characters of 255 from the data, the mean number of characters was 521 per post (SD = 397.88, Mdn = 426.5). In practice, influencers are advised to be short and precise and to put important information at the beginning of captions (Influencer MarketingHub 2019; Jackson 2017). Thus, we are confident that a character count of 255 is sufficient to compute a reliable metric for the intensity and direction of emotions in a post.

Our study provides the groundwork for future research on the use of ES in influencer advertising. The study presented here helps focus on the effects of overall emotionality, its dimensions in sponsored posts, its interactions with the influencers’ number of followers on social media engagement, as well as some areas for future research. We look forward to exploring these and other possible avenues in our future work.

Acknowledgments

This article is based on essay four of Gross’s (2020) doctoral thesis Thumbs Up for Brands: Influencer Marketing in the Era of Social Media. We refer the reader to Gross (2020) for more detailed discussions.

Disclosure statement

On behalf of all authors, the corresponding author states that there is no conflict of interest.

Data availability statement

Due to the nature of this research, participants of this study did not agree for their data to be shared publicly, so supporting data are not available.

Additional information

Notes on contributors

Jana Gross

Jana Gross (PhD, ETH Zurich) is an assistant professor, KEDGE Business School.

Zhiying Cui

Zhiying Cui is a doctoral student, Department of Management, Economics, and Technology, ETH Zurich.

Florian von Wangenheim

Florian von Wangenheim (PhD, University of Mainz) is a professor, Technology Marketing Group, and head, Department of Management, Technology, and Economics, ETH Zurich.