Analyzing the in ﬂ uence of celebrities ’ emotional and rational brand posts

Purpose – The purpose of this paper is to compare two brand posts with the different content created by a celebrity (emotional content vs rational content) on Instagram and their effects on users ’ willingness to use of ﬂ ine word of mouth (WOM) and electronic WOM (eWOM). The research model also consists of product involvement as the moderator. Design/methodology/approach – Based on the results of the pretest stages, the study included two stimuli, and respondents were presented with two different brand posts (i.e. two manipulated pictures and texts on the Instagram frame). A two-group comparison (positive emotional brand post vs negative rational brandpost)between-subjectsexperiment( n = 214) was conducted. Findings – The results indicate that WOM and eWOM are more affected by a celebrity ’ s emotional brand post than a celebrity ’ s rational brand post. Furthermore, both types of WOM are more affected through high product involvement enhanced by a celebrity ’ s rational brand post than through high product involvement boosted by a celebrity ’ s emotionalbrandpost. Practical implications – Managerial implications for social media marketing and Instagram celebrity- based branding are provided. Practical implications are also provided in the form of evidence of how the impacts of two different brand posts on positive of ﬂ ine WOM and eWOM are affected differently through the moderation of product involvement. Originality/value – The research argues for theoretical implications for the marketing literature on celebrity endorsements. The study also tests one moderating effect on the relationship between brand post contentandWOMandeWOM.

Analizando la influencia de los mensajes emocionales y racionales de los famosos sobre las marcas Resumen Prop osito -El prop osito de este trabajo es comparar dos posts de marcas con diferente contenido creado por una celebridad (contenido emocional vs. contenido racional) en Instagram y sus efectos en la disposici on de los usuarios a utilizar WOM y eWOM. El modelo de investigaci on tambi en incluye la implicaci on del producto como moderador.
Diseño -Sobre la base de los resultados de las etapas de prueba previa, el estudio incluy o dos estímulos, y a los encuestados se les presentaron dos publicaciones de marca diferentes (es decir, dos im agenes y textos manipulados en el marco de Instagram). Se realiz o un experimento entre sujetos (n = 214) de comparaci on de dos grupos (post de marca emocional positivo frente a post de marca racional negativo).
Conclusiones -Los resultados indican que el WOM y el eWOM se ven m as afectados por el post emocional de la marca de un famoso que por el post racional de la marca de un famoso. Adem as, ambos tipos de boca a boca se ven m as afectados por la alta implicaci on del producto potenciada por el post racional de la marca de un famoso que por la alta implicaci on del producto potenciada por el post emocional de la marca de un famoso.
Implicaciones pr acticasse ofrecen implicaciones de gesti on para el marketing en las redes sociales y el branding basado en los famosos de Instagram. Las implicaciones pr acticas tambi en se proporcionan en forma de evidencia de c omo los impactos de dos publicaciones de marca diferentes en el WOM positivo y el eWOM se ven afectados de manera diferente a trav es de la moderaci on de la implicaci on con el producto.
Originalidad -La investigaci on aporta implicaciones te oricas para la literatura de marketing sobre el patrocinio de los famosos. El estudio tambi en prueba un efecto moderador en la relaci on entre el contenido de los posts de marca y el WOM y el eWOM.

Introduction
Currently, the growing popularity of social media networks empowers individuals to acquire large audiences of up to several million people on these platforms. Social network platforms create hubs of brand communities that congregate users who contribute to newsfeeds and share brand information by tagging friends (Wang, 2021). Instagram, as one of these popular social networks, is the fastest growing, recently reaching one billion four hundred and seventy-eight million active users (Statista, 2022). Instagram has enabled users to take photos, use filters, share the photos with followers who can like and comment on them (Kim et al., 2017) and provide other templates (e.g. stories and live videos) to improve the user experience. The venue format of platforms such as Instagram enables brands to create owned media and to engage with their followers by constructing posts using visual and textual modalities (Rietveld et al., 2020). Brands can reach a large audience at a relatively low cost compared to paid media activities such as advertising. Nike, for example, has attracted 83 million followers over the last seven years (at the time of writing) and published 700 unique posts leading to different levels of likes and comments per post (Rietveld et al., 2020).
Realizing what drives customer engagement across these posts is significant for brands because customer engagement initiatives have a positive effect on financial returns (Beckers et al., 2018). In this regard, some former studies (King et al., 2014) called for more investigation of the visual modality. The impacts of the visual modality have typically been investigated in advertising (Akpinar and Berger, 2017), and preceding research in social media predominantly concentrated on the effects of the textual modality (Lee et al., 2018). However, previous scholarly research has examined the effects of the visual modality on social media (Rietveld et al., 2020).
In these investigations, message content arises as a crucial driver of engagement behavior. While motivations for customer engagement may differ (Berger, 2014), customer engagement is at least partly determined by exposure to message content. Recent investigations on social media propose that the distinction between emotional and informative appeals offers a significant lens for studying the impact of message content on engagement behavior (Akpinar and Berger, 2017;Lee et al., 2018). Akpinar and Berger (2017) found that online video advertisements with robust emotional appeals are more presumably to be shared, while online advertisements with informative appeals drive brand evaluations and purchases. Similarly, Lee et al. (2018) found that regulating emotional and informative appeals in the textual content on Facebook results in different brand-related consequences.
In addition to using emotional and informative appeals in the content on Instagram, companies often use traditional celebrities as endorsers to increase consumers' attraction and engagement with advertising posts. Companies typically expect consumers to become empathic through the high popularity or attraction of a celebrity and then feel good about the products being recommended (Lin, 2011). However, in recent years, using celebrities on social networks has been faced with hesitation such that their place has been taken by influencers in corporate advertising. This is especially true on Instagram, where celebrities have been less prominent than influencers in recent years (Jin et al., 2019). Online influencer marketing has become an integral ingredient of brands' marketing strategies; however, marketers lack a sufficient understanding of its scope, effectiveness and possible threats (Leung et al., 2022).
To perform a successful marketing strategy for companies regarding how to use celebrities on social networks, especially on Instagram, marketing experts need to identify what type of content (emotional and informative appeals) works best for the specific company. To achieve this purpose, we test two promotional brand posts with different content that can mostly have different outcomes resulting from the brand posts shared by celebrities on the social media platform. The two contents used in brand posts are known as emotional and rational content. Emotional and rational contents give the brand a powerful indication and promote category-based processing. In this research, we conduct an experiment comparing how users react differently to types of promotional brand posts (designed in the form of emotional content and rational content on Instagram) shaped by one celebrity endorsing the same product. We expect these brand posts to have different outcomes on consumer behavior, such as willingness to spread positive offline word of Emotional and rational brand posts mouth (WOM) and electronic WOM (eWOM). In addition, an analysis is made of the level of involvement of the user with the type of product endorsed in the advertising posts.
To the best of our knowledge, this is the first attempt to compare the effects of these two brand posts created by the same celebrity (emotional post vs rational post) on an Instagram frame. Novel findings are also provided in the form of evidence of how the effects of these brand posts on positive offline WOM and online WOM are affected through the moderation of product involvement. Based on the obtained data from Business 2 Community (2018), 74% of consumers recognize WOM as a crucial influencer in their purchasing decision, 68% of people trust online opinions from other consumers, which places online opinions as the third most trusted source of product information, and 81% of people express that they are affected by what their friends share on social media. Thus, companies that have a strong social media presence online are usually successful with WOM marketing. The results of the research contribute to digital marketers finding out before launching a marketing campaign by a celebrity whether using brand posts with emotional content is more effective for their consumers or using brand posts with logical content.
2. Literature review and hypothesis development 2.1 Celebrities' influence on social media Consumers are progressively using social media to meet information on which to find their decisions (Casal o et al., 2020). Numerous advertisers have appeared as influential members of online societies and have been publicized to be a source of advice for other consumers (Thakur et al., 2016). Studies concentrating on how to target influential in social networks explore the dynamics of user influence in terms of retweets and mentions. They remark that most advertisers can have an important influence on a diversity of topics and that people with more associates might exert an even better influence on information dissemination (Zhang et al., 2016).
Advertiser influence in social networks can help to explain the successful diffusion of innovative ideas and new products. For example, commercials for products or services are often presented by a celebrity, which is a very good reason for this. Due to their great popularity and fame with the people, celebrities have many fans and can change the way companies work in their favor (Jia et al., 2020). Many TV and movie stars have appeared in advertisements for various products and have said positive things about the brands. This has a major psychological effect on consumers because these celebrities are dependable. For example, if a company is not very well identified by the public, it can use a celebrity to endorse its products and spread the word (Cleverism, 2019). People may distinguish the celebrity very well and trust their opinion on the matter when they say something. As a result, people will start purchasing products that they would not have thought about otherwise. These celebrities have influenced consumer behavior just by sponsoring some products. There are many examples of the effect of celebrities' endorsements on consumers' behavior, which can underline their effective role as opinion leaders (Cleverism, 2019).
Over the years, in addition to influencers, celebrities have determined how to communicate directly with fans on social networks. Platforms such as Twitter and Instagram have provided users a closer look at their favorite celebrities, yielding them direct access to their world. For instance, celebrities such as Selena Gomez and Chrissy Teigen have operated their platforms to speak openly with fans about significant subjects, whereas stars such as Kim Kardashian, Kylie Jenner and Justin Bieber have constructed their empire on social media. These celebrities have the power to influence millions of individuals worldwide with a single post, and they have mastered doing it (Usmagazine, 2020).

Types of brand post contents
On social media, message content appears to be a crucial driver of engagement behavior (Rietveld et al., 2020). Although motivations for customer engagement may differ (Berger, 2014), consumer engagement is at least partly determined by exposure to message content. Mainly, there are two types of content shared on social media, emotional and rational (Manch on et al., 2014). Consistent with Lim et al.'s (2018) statement, rational and emotional content are expected to have a different effect on the outcome of consumer behavior with regard to customers' interaction with the brand.
Emotional content applied in advertising can be classified into positive and negative according to its valence (Taute et al., 2011). The main negative emotions can include guilt, fear and shame (Keshari and Jain, 2014) or anger, sadness, frustration and hopelessness (Kim and Franklin, 2015). On the other hand, positive emotional content appears when the message on social media includes patriotism, affection and nostalgia (Panda et al., 2013) or humor, joy, emotions, holiday mood, entertainment, excitement and emoticons (Keshari and Jain, 2014).
Rational messages online are intended to be intellectually processed (Dolan et al., 2019). Rational/informational contents concentrate on consumers' practical, functional or utilitarian requirements for the product or service and highlight features of a product or service and/or the advantages or reasons for possessing or consuming a specific brand (Grigaliunaite and Pileliene, 2016). The content of these messages highlights facts, learning and the logic of persuasion. Rational content is inclined to be informative, and advertisers apply them usually to make an effort to convince consumers that their product or service has a specific characteristic(s) or provides a particular benefit that contains their requirements (Grigaliunaite and Pileliene, 2016). Their purpose is to encourage the target audience to purchase the brand because it is the best available or performs a better job of meeting consumers' requirements (Belch and Belch, 2004). Advertising can be reflected as rational content advertising if the advertisement comprises one of these information cues: function, material, quality, the price of the product or service, purchasing time and place, packaging or any research data about the product (Keshari and Jain, 2014). If the advertising encompasses none of the above information, it is considered emotional content advertising.

Word of mouth
WOM can be recognized as any positive or negative statement constructed by potential, actual or earlier customers about a product or a company, which is made obtainable through offline or online channels (Hennig-Thurau et al., 2004). Although the nature of WOM behavior is consistent whether based offline or online, there are differences. eWOM mostly includes written communication (whereas offline WOM is mostly spoken) and may occur in one-to-one, one-tomany or many-to-many forms (offline WOM mostly occurs in one-to-one form) (Karjaluoto et al., 2016). The massive upsurge in the utilization of the internet has transformed traditional WOM into eWOM (Reyes-Menendez et al., 2019). Electronic WOM is more exposed to a ripple effect and may be more simply controlled by companies compared to offline WOM (Huang et al., 2011). Furthermore, Lastovicka and Sirianni (2011) recommend that consumers with beloved objects may be selective in their offline WOM behavior and engage in WOM only in a certain context or only with chosen individuals. Consequently, it is logical to differentiate between offline WOM and eWOM. Both behaviors (WOM and eWOM) are very relevant for firms.

Effect of celebrities' brand posts
Prior studies have recommended that when someone endorses a product on social media, the audience expands positive reactions such as spreading positive WOM (De Veirman et al., 2017) if they believe that the product's review is valid (Spry et al., 2011). However, if the Emotional and rational brand posts review is perceived as false and invalid, the audience will be expected to develop a negative reaction toward the brand and the endorser (Cheung et al., 2009). This proposes that the quality of the social media content is appraised in regard to the perception of the endorser, as well as relevance (Djafarova and Rushworth, 2017). In contrast to celebrities, some social media opinion leaders, such as influencers, promote a product or service in an authentic way (Schouten et al., 2019), and typically, it is done by constructing personal relationships with their audiences on social media channels (Lou and Kim, 2019). Previous investigations have demonstrated that celebrities are perceived as less credible than such opinion leaders because of people's skepticism (De Veirman et al., 2017). Consumers who are very skeptical about the truth of advertising claims are more responsive to emotionally appealing ads than those peppered with information (UW News, 2005). This means that their endorsements are more likely to activate the consumers' persuasion knowledge who may recognize content advertised by a celebrity as a persuasion tactic (Hwang and Jeong, 2016). The use of emotions in advertising is ubiquitous and effective, and consumers can resist persuasion by focusing on the available emotions in advertising, which is clearly worthwhile (Lewinski et al., 2016). Therefore, it is expected that celebrities will be more effective in spreading the WOM of the brand post published when using emotional promotional content than when using rational content, as this content, in addition to presenting celebrities' endorsed products, potentially avoids skepticism and resistance (Lu et al., 2014).
Previous studies have also proven that celebrities have less trustworthiness and credibility on social networks (Jin et al., 2019;De Veirman et al., 2017); hence, emotional content by celebrities may be considered more direct and appropriate for giving recommendations to consumers. Consumers are more likely to talk about content that can establish an emotional (vs informational) connection with them (Botha and Reyneke, 2013). It has been found that emotions play an influential role in motivating customers to spread positive WOM (Berger and Milkman, 2013). In addition, advertising can further reflect emotions in its content via affective messages, visuals, audio and celebrity endorsements (Rosenkrans, 2010). The use of emotional appeals in advertisements can boost the likelihood of sharing on social media by activating arousal (Berger and Milkman, 2013). Therefore, it is also expected that celebrities will be more effective in the audience's positive behaviors, such as spreading positive offline or online WOM, when using emotional promotional content than when using rational content. Thus, we propose the following hypothesis: H1. Celebrity brand posts with emotional and rational content have a positive effect on positive WOM; however, emotional brand posts have a stronger effect on positive WOM than rational brand posts.
H2. Celebrity brand posts with emotional and rational content have a positive effect on positive eWOM; however, emotional brand posts have a stronger effect on positive eWOM than rational brand posts.

Moderating role of product involvement
Product involvement has been defined as the individuals' amount of interest in a product or the relevance assigned to a purchase situation (Zaichkowsky, 1985). Involvement depends on personal factors (e.g. needs, importance, interest, values) and situational features (e.g. differentiation of alternatives) and may be related to different outcomes (e.g. relative importance of the product category, perceived differences in product characteristics and preference for a specific brand) (Zaichkowsky, 1986). Thus, the degree of product involvement may vary from person to person; accordingly, user behavior may also differ meaningfully. Consumers with high product involvement typically seek to maximize their satisfaction through a conscious decision-making process (Laurent and Kapferer, 1985). In this manner, involved consumers search for and apply more information to make their decisions and are more interested in learning about products. On the other hand, less involved consumers tend to simplify their choices by using other risk-reduction strategies (Lockshin et al., 2006). In the advertising context, individuals with low involvement with a product often pay more attention to emotional messages and are unwilling to make great efforts to process any information received (Petty and Cacioppo, 1986). On the other hand, it has been noted that greater involvement motivates the adoption of a more attentive state of mind, which allows the consumer to better process information (Yoo et al., 2004). Thus, users who are more involved with the promoted product category pay less attention to emotional messages and process them less intensively (Baker and Lutz, 2000). In addition, when individuals are involved, they feel less irritated when provided with rational messages. However, when these messages are generated by celebrities who have low trustworthiness on social networks, then it is logical that this degree will be somewhat diminished and motivate individuals with high product involvement less. Feeling less trusting in the advertiser can question the reliability of a product (Lu et al., 2014). In addition, Wu and Wang (2011) suggested that consumers with high involvement (vs low involvement) are most likely to apply central route information processing and are more willing to carefully elaborate on eWOM content to obtain additional product information. Hence, when there is a rational brand post created by a celebrity, it can increase the interest and attention of consumers with high product involvement, which results in more positive WOM. Therefore, the following hypotheses are proposed: H3. The effect of a celebrity's brand posts on positive WOM is moderated by consumer involvement, such that highly involved consumers moderate the effect of a rational brand post (vs an emotional brand post) on positive WOM better than less involved consumers.
H4. The effect of a celebrity's brand posts on positive eWOM is moderated by consumer involvement, such that highly involved consumers moderate the effect of a rational brand post (vs an emotional brand post) on positive eWOM better than less involved consumers.
The proposed research model summarizes the set of hypotheses developed (Figure 1). These hypotheses articulate the means to achieve the proposed objectives of this research.

Research methodology 3.1 Participants
The study was introduced to female students from a large international university in Thailand using convenience sampling. Participants took part in a lab session where the study was conducted. The research focused on women only as the majority of influencers tailor to a female audience (Gannon and Prothero, 2018). Moreover, it focused on women between the ages of 18 and 40, as this constitutes the most common Instagram demographic, with 64.1% of women between the ages of 18 and 40 using Instagram compared to only 22.6% of women over 40 (Pew Research Center, 2018). Some control questions were included at the beginning of the survey to confirm that the participants knew the celebrity, her features and the kind of content she posts (e.g. her age and category of products she promotes). Participants who responded incorrectly were automatically excluded from the study. Finally, 214 female students (M age = 29.33 years) participated in a two-cell, one-way (celebrity advertising post: positive emotional vs rational) between-subjects design for a Emotional and rational brand posts cash gift ($5) as a token of appreciation for their participation in this survey and were randomly assigned to one of the two conditions.

Design and manipulation
An experiment was designed to test the research hypotheses. We focused on a popular fashion celebrity on Instagram. Davika Hoorne (also known as Mai Davika) was selected because she represents a celebrity figure with high recognizability and attractiveness. She is a young Thai woman who had previously enjoyed celebrity status (e.g. as an actress and model). She has a high number of Instagram followers, which helps not only ensure ecological isomorphism and a reasonable level of external validity but also controls the level of quantitative Instagram popularity across conditions. In the celebrity's account, we searched for both emotional and rational advertising posts that were determined based on the visuals and the textual content of the advertisements to stimulate individuals. We selected four promotional posts containing four pictures that completely fit with the definitions of the brand posts (containing positive and negative emotional content and positive and negative rational content) as a way to influence potential women fashion customers. The pictures were manipulated with images of the celebrity displaying she was promoting the products (face serum) of a luxury cosmetics brand (L'Or eal). Then, we added the texts in the caption of the advertised posts corresponding to the context of manipulated post pictures. In the next stage, we performed a pretest.

Pretest
To ensure the accuracy of the brand posts, we asked a range of individuals (colleagues and PhD researchers) to check the brand posts for validity and clarity. After ensuring the accuracy of the brand posts, we ran a pretest (86 students of the Faculty of Advertising) including two phases. We entered the first phase to determine whether participants understood the contents of the brand posts correctly and realized the variance between different contents of the brand posts. For each brand post, participants showed the degree to which brand posts displayed and expressed conditions (1 = extremely disagree/7 = extremely agree). The results showed that  Figure 2 shows that the final advertising posts formed by the celebrity (positive emotional post and negative rational post) were confirmed based on the results of the pretest stages.

Procedure and measure
All participants were presented with two stimuli. They were given brand posts with different contents (i.e. positive emotional content and negative rational content) on an Instagram frame. Then, each of them completed a questionnaire. The scales for willingness to engage in positive  Carroll and Ahuvia (2006), and the scales for willingness to engage in positive eWOM (three items) were adjusted from the scales applied by (Karjaluoto et al., 2016). The items of eWOM (i.e. I "talk up" this brand in online environments; I give this brand tons of positive word of mouth on the internet; I try to spread the good word about this brand on the internet) differ from offline WOM (i.e. I have recommended this brand to lots of people; I "talk up" this brand to my friends; I try to spread the good word about this brand; I give this brand ton of positive word of mouth advertising) in several ways. The most obvious difference is that it is obtained online, while WOM is conducted through a face-to-face communication process, such as during meetings and telephone conversations (Doma et al., 2015). Online discussion forums, blogs and emails are most often used for eWOM. The moderator-product involvement was measured by five items (i.e. uninterestedinterested, not involved-highly involved, of no concern-of concern to me, unimportantimportant and irrelevant-relevant) adapted from Wang et al. (2012) and Zaichkowsky (1985) assessing users' level of the product category. The descriptive items were measured using seven-point Likert-type scales (1 = "strongly disagree" and 7 = "strongly agree"). Following the standard process, to generate a comparison between consumers based on their involvement with the product (Jaccard and Wan, 1996;Jin, 2009), the participants were divided into low-and high-involvement groups based on a median split (M = 4.68). Around this mean, some cases were eliminated (61/2 standard deviation [SD]; García et al., 2008). The collected data was analyzed using the IBMSPSSAmos 23.0 statistical program.

Common method bias
To avoid the common method variance of the measurement constructs, procedural and statistical examinations to control common method bias (variance) were applied. Common method variance is defined as that which is attributable to the measurement rather than to the construct of interest. Variance can affect outcomes because of response biases, such as halo effects, social desirability, acquiescence, leniency effects or yes and no answers (Podsakoff et al., 2003). Thus, we collected data through face-to-face surveys and a question randomization option for each respondent in a shuffling approach. Furthermore, we carefully constructed the items to avoid ambiguous or unfamiliar terms and vague concepts; when such concepts were applied, the respondents were provided with examples; the questions were given in a simple, specific and concise manner, with double-barreled questions, and complicated syntax was avoided (Tourangeau et al., 2000). In terms of the statistical measure to control for common method variance, we used Harman's single-factor test with an unrotated factor solution through EFA. The result designates an explained variance of 24.7%, which is relatively less than the threshold of 50% recommended by Podsakoff et al. (2003).

Reliability and validity of the constructs
A confirmatory factor analysis was performed with all factors for the theoretical research model. The goodness-of-fit statistics for the measurement model suggest a reasonable fit to the data (Hu and Bentler, 1999) as follows: x 2 (29) = 49.936, p = 0.001, the comparative fit index (CFI) = 0.984, the non-normed fit index (NNFI) = 0.975, the standardized root mean squared residual (SRMR) = 0.027 and the root mean square error of approximation (RMSEA) = 0.068. The results of the reliability and validity of the factors (Table 1) showed that all relevant loadings are substantial and highly significant. Item factor loadings for the constructs were all greater than 0.60, and Cronbach's a values were close to or exceeded the cutoff value of 0.70 (Bagozzi and Yi, 2012). The model also passed the tests of convergent and discriminant validity (Table 1), as average variance extracted (AVE) values exceeded the cutoff of 0.50, and the squared AVE values were larger than the correlations shared by the respective paired constructs (Fornell and Larcker, 1981).

Manipulation check
As a manipulation check, we asked respondents to indicate whether the brand post is perceived as emotional content or rational content on a seven-point Likert scale (1 = totally disagree; 7 = totally agree). The manipulation of brand posts was successful, with perceptions of brand posts being higher in the emotional condition (M = 5.73, SD = 1.25) than in the rational condition (M = 4.49, SD = 1.63), t(212) = 19.12, p <0.01, d = 2.47.

Hypothesis testing
Two sets of analyses were used to test the theoretical model. H1 and H2 were analyzed using linear regression and analysis of variance (ANOVA) techniques to evaluate the direct effects of variables on offline WOM and online WOM. H3 and H4 were analyzed using linear regression/ANOVA and the PROCESS macro (Hayes, 2013; Model 1).

Direct effects of the brand posts
The first linear regression and ANOVA were conducted using two independent variables (emotional brand post/rational brand post) and a single dependent variable (WOM). The first hypothesis assumes that emotional post and rational post have a positive effect on WOM (H1: b = 0.56, t = 3.24, p < 0.05; b = 0.32, t = 2.89, p < 0.05). The results indicated that the emotional post has a greater effect on individuals' encouragement to spread eWOM than the rational post [F(1,212) = 10.46, p < 0.02, M em = 5.34, M ra = 4.72], providing support for H1. As expected, the results provided support for H2, which stated that brand posts with emotional and rational content have a positive effect on eWOM (H2: b = 0.64, t = 3.57, p < 0.05; b = 0.38, t = 3.06, p < 0.05). The outcome also shows that the emotional post has Emotional and rational brand posts a higher effect on individuals' encouragement to spread eWOM than the rational post [F(1,212) = 11.23, p < 0.02, M em = 5.58, M ra = 4.61], indicating that H2 is supported.

Moderation analysis
The first moderation test was conducted on the PROCESS macro (5,000 bootstrap resamples) using one independent variable each time (emotional post/rational post), one moderator (product involvement: high vs low) and one dependent variable ( The second moderation test was run similarly to the first moderation test, and we only replaced eWOM as the dependent variable. The results also provided support for H4. The interaction of product involvement and brand post condition had a significant effect on eWOM (B = 0.57, SE = 0.28, p = 0.01). When the level of involvement was high, there was a significant effect of brand posts on eWOM (B = 0.42, SE = 0.21, p = 0.02). In addition, when the level of involvement was low, a significant effect of the level of involvement on PWOM was found (B = 0.23, SE = 0.19, p = 0.01). A further analysis also shows that the rational post increases eWOM among individuals highly involved with the product but decreases eWOM among individuals with low product involvement [B = 62, t(1,211), p < 0.01, M high = 5.63, M low = 2.47]. In contrast to a rational post, an emotional post has a greater effect on eWOM among individuals less involved with the product than among individuals with high product involvement [B = 42, t(1,209), p < 0.01, M low = 5.09, M high = 2.14]. The results of the two moderation analyses are presented in Table 2 and Figure 3.

General findings related to hypothesis tests
Our paper documented the results of the effects of two brand posts (emotional post vs rational post) created by the same celebrity on two types of consumer behavior (WOM and eWOM). The paper also detailed the results of the effect of product involvement on the relationship between brand posts and behaviors. Table 3 presents the general results of testing the research hypotheses.

Discussion and theoretical contributions
The current research helps companies and brand owners determine how to build successful marketing campaigns on Instagram using traditional celebrities. The present research is one of the first attempts to establish a theoretical framework that explains the effects of social media celebrities by providing some evidence that audiences perceive different emotional brand posts and rational brand posts created by these celebrities differently.
Previous studies have examined the effect of emotional and rational content on other consumer behaviors and in different contexts (Manch on et al., 2014;Lim et al., 2018) but not on WOM and in the context of celebrities' brand posts. In addition, to explain differences in perception more specifically, the experiment drew upon product involvement theory in examining why people connect more strongly with rational content created by celebrities than with their emotional content. The results of the research showed that the celebrity's emotional brand post better encourages audiences to spread positive WOM and eWOM compared to the celebrity's rational brand post. The findings also revealed that compared to lowly involved consumers, highly involved consumers better moderate a celebrity's rational brand post on WOM and eWOM than a celebrity's emotional brand post.
The findings support our expectations that a celebrity's emotional brand post has a greater effect on positive behaviors such as positive WOM and eWOM than a celebrity's rational brand post. This contributes to the prior research, which, due to a lack of sufficient expertise, identifies celebrities as less aware of the features of the products and services they 6.11*** 7.15*** Notes: *p < 0.05; **p < 0.01; ***p < 0.000 Emotional and rational brand posts advertise (Schouten et al., 2019). In addition, consumers find such ads less credible and may express more skepticism about them when they are shared by a traditional celebrity. Another reason that highlights the role of emotional posts shaped by celebrities can be that emotional advertisements can have a recall advantage and a positive relationship with brand trust. This means that such emotional advertisements have more influence on customer attitudes through the theme of the message. Rational content includes facts such as product features, information and specifications (Dolan et al., 2019), and publishing brand posts including such facts on social media is better endorsed by knowledgeable individuals or experts. Peer reviews and suggestions are increasingly influencing consumers' buying decisions; therefore, marketers have made obtaining positive WOM the focal objective of marketing activity (Chen and Xie, 2008). However, prior research is very limited in showing how types of brand post content affect consumers' willingness to engage in positive WOM in offline (Kim et al., 2020) or online (Wu and Wang, 2011) contexts. This study extends the literature by showing that an emotional brand post shared by a celebrity has a strong positive effect on both types of WOM. However, the effect is higher for online WOM (M = 5.58).
The moderating role of product involvement has not been widely studied in the context of celebrities' brand posts and WOM, although the concept of message contents (rational content vs emotional content) itself is conceptualized from the long relationship perspective (Wu and Wang, 2011). Some previous literature has separately examined the moderating effect of product involvement between emotional and rational content and consumer behaviors in different contexts. For example, the moderating effect of product involvement has been tested on the relationship between positive and negative emotions and satisfaction in the context of restaurants (Calvo-Porrala et al., 2018), and the moderating effect has been examined on the relationship between rational content and cognitive response in a printed advertisement context (Wu et al., 2017). Nevertheless, the empirical findings of our research demonstrate the mediating role of product involvement in determining the effects of brand posts created by celebrities on WOM and eWOM. These robust findings across different dimensions of consumer behavior (offline WOM and online WOM) verify the power of product involvement in explaining consumers' reactions to brand posts. Our study is among the first to confirm that the more highly involved consumers are, the higher the relationship between a celebrity's emotional post and WOM will be. Although for low involvement products, emotional ads created by celebrities have a more significant effect, for high involvement products, rational ads are more significant. To a large degree, these alterations display the wholly opposite viewpoints accepted by each advertising post, with implementation elements in rational ads revolving mostly around objectivity, functionality and utilitarianism, contrasted with emotional advertising elements that are considered more by subjectivity, emotionalism and value expressiveness. However, these advantages will be alleviated when the celebrity is introduced as an advertiser. This also happens due to the lower credibility of celebrities on social networks. The last theoretical contribution can point to the results gained from the pretest stage, where the positivity of the emotional brand post and the negativity of the rational post brand attracted more audience attention. These findings are consistent with the results of previous research; for example, Kim and Franklin (2015) specified that many scientific investigations view positive emotions as the desired result, which are related to positive expectations such as hope, faith, courage and trust. It could be acknowledged that such an outcome emphasized in the advertisement can encourage consumers to purchase the promoted product. In addition, advertisements consistent with positive emotions create good feelings and positive relations for consumers regarding the advertised brand (Casal o et al., 2021;Panda et al., 2013). Former research (De Pelsmacker et al., 2002) also proposes that a negative context, or a context that evokes a negative mood, leads to better information processing. Moreover, the feelings as information theory propose that when people are in a negative mood, they will look for stimuli that could alter this state (Worth and Mackie, 1987). Hence, when people are in a negative mood, information processing takes place more carefully.

Managerial implications
Based on the results of the research, several recommendations for the companies and brand owners have been provided. The first is that in the case of using a celebrity to promote brands/ products on social networks, especially on Instagram, shaping posts with emotional content, especially with positive content, promoted by the celebrity can engage more consumers to spread positive WOM. For example, using moods such as happiness or excitement (i.e. high arousal) has a positive impact (Beaudry and Pinsonneault, 2010) and can contribute to more consumers' encouragement to extend positive WOM. An emotional brand post, which is very well designed, can generate a positive change toward a product and not bring contra-productive results.
Because the consumers' interest is lower to look for the rational content of branded products advertised by celebrities on social networks such as Instagram, the research recommends that in case of using the celebrities, the companies can consider repetition of advertising posts or spreading hashtags. Furthermore, the brand owner should, by displaying more truthfulness, credibility, accuracy and usefulness of promotional product/brand posts, engage more individuals with such rational content. They can use this mechanism by sharing the posts on the more reliable accounts or the dependable pages of other popular individuals on social networks. Emotional brand posts formed by celebrities that depict enjoyment and happiness affect lowly involved customers, dramatically resulting in more WOM. On the other hand, highly involved customers are prone to pay more attention to rational messages. Messages depicting logic such as information and quality are more important for highly involved consumers. However, in the case of using a celebrity as an advertiser, the truthfulness and reliability of such advertising content will be somewhat declined, and they are less influential and do not create a special influence on the individual's engagement with the posts. Thus, online brand managers can test a mixture of contents.

Conclusions
A summary of the investigation findings along with the theoretical and managerial implications is provided in Table 4. Emotional and rational brand posts 6. Limitations and future research This research has limitations that pave the way for the development of future research lines. First, it only concentrated on a single product category and brand. The concentration was limited to fashion branding because it is becoming a pervasive phenomenon in social media marketing, chiefly on visual platforms such as Instagram. It is essential to note that numerous industries or social media apps have affordances and spirits different from what Instagram presents. Second, the study was directed with just one celebrity who, despite being very famous in Thailand (the country of origin of the participants), may have some particularities and special characteristics in comparison to other similar celebrities. Consequently, to generalize the results, future research should replicate the study with other celebrities. Third, this research included one moderating variable to help better understand the phenomenon under study. However, future research should consider other variables that may serve to elucidate the underlying mechanisms of these types of celebrity-consumer interactions. Finally, our research sample was made up of women between the ages of 18 and 40; however, future research should consider a gender balance with different age samples to simplify the findings. Celebrities' emotional brand post (vs celebrities' rational brand post) encourages better audiences to spread positive WOM and eWOM -Due to a lack of sufficient expertise, celebrities are less aware of the features of the products and services they advertise, and consumers recognize such ads as less credible -To promote brands/products by celebrities, shaping posts with emotional content, especially with positive content (e.g. happiness or excitement) is more effective Highly involved consumers (vs lowly involved consumers) celebrity's rational brand post on WOM and eWOM -In the case of using a celebrity as an advertiser, the truthfulness and reliability of advertisements with informational content will be somewhat declined for highly involved consumers -To control the engagement of highly and lowly involved customers, a mixture of contents can be tested