Parasocial interaction can be described as the relationship between personas (e.g., actors, presenters, or celebrities) in the mass media and their audience. This intertwined relationship influences the audience's perception of the persona’s behavior (Zheng et al., 2020). There are several definitions of parasocial interaction. In the marketing sector, Labrecque (2014) defines PSI as an imaginary reciprocal relationship between an audience and a particular persona. Such relationship is developed if the audience continues to follow the persona’s activities on the media. According to Horton and Wohl (1956), parasocial relationship and parasocial interaction have the same meaning, which is why we adapted some supporting theories of PSI based on the concept of parasocial relationships in this study.

PSI provides a relevant framework to understand the one-sided relationship between celebrities and their followers, in which one party (namely the celebrity or influencer) is not aware of it (Kim & Song, 2016). An example of this could be a woman who admires a celebrity (whom she always watches on television) and tries to look like her in order to become part of her community. These relationships in traditional media settings are in line with those in digital media sites between social media users and influencers: for instance, a subscriber’s admiration towards a social media influencer makes her build a close relationship between both. This is supported by the findings of Lee and Watkins (2016), who point out that PSI provides a conceptual framework to describe the relationship between celebrities and their followers on online media platforms, such as that between vloggers and social media users. This relationship is developed when users decide to subscribe to the YouTube channel of vloggers, follow everything they share on social media, and interact with them through comments. Vloggers who have a good relationship with their subscribers will find it easier to persuade them to follow their vlogs.

Everyone processes information differently depending on their own characteristics. In this regard, the Elaboration Likelihood Model (ELM) theory suggests that individuals’ motivation and abilities influence the way they understand information, which will ultimately affect their attitude towards the information received. Incoming information can be processed through either the central route or the peripheral route (Shimp & Andrews, 2017). When it is processed via cognitive responses to the primary message, individuals are using the central route, while when it is processed only by focusing on supporting elements (e.g., pictures or image of the persona) but not on the primary message, individuals are employing the peripheral route.

Message recipients who prefer peripheral cues process information based on supporting attributes that act as the source of attraction. One of those supporting attributes is source attractiveness. It indicates that the image of a persona (e.g., an actor, presenter, or celebrity) can be the source of attraction for the audience to understand the message being delivered. Thus, actors or celebrities who regularly appear in the media often have an attractive appearance that stirs up feelings of admiration in the audience or a desire to look similar to them. In addition, such attribute encourages the formation of an imaginary relationship between personas and their audience. According to Rubin and McHugh (1987), the development of parasocial interaction relationships is influenced by the perceived attractiveness of performers.

Such constructed attractiveness is also a factor considered by vloggers when creating product review contents on YouTube. They pay great attention to their physical appearance and attitude, which will be visible on their YouTube channel, as this is expected to become a source of attraction for their viewers to further watch the video of the product that is being reviewed (Rosaen et al., 2019). Consequently, the greater the appeal of a persona, the higher the chance that such persona and his or her viewers will interact and the greater his or her ability to influence them.

McCroskey and McCain (1974) divide attraction into three categories: task attractiveness, physical attractiveness, and social attractiveness. Physical attractiveness refers to the physical appearance of a persona; for example, beauty vloggers with attractive facial features or an attractive physical appearance. Social attractiveness can be described as a persona’s personality that makes him or her likable. A beauty vloggers’ positive and pleasant personality can make their subscribers to like them and to spend more time watching every content they upload on their YouTube channel. Task attractiveness is the ability of individuals to reflect the desired personality. According to Klimmt et al., (2006), the type of attraction that favors parasocial interaction relationships varies depending on the media used. For instance, as mentioned by Lee and Watkins (2016), physical attractiveness and social attractiveness have a higher influence on the development of parasocial interaction relationships in online media sites.

Aral and Walker (2014) state that attitude homophily refers to individuals’ tendency to bond and interact with other individuals who are similar to them in terms of education, social status, among other characteristics. When people feel that they share some similarities with others, such as same desires and interests in a particular field, they will more likely interact with them. For example, if a subscriber feels that she has things in common with beauty vloggers in terms of keeping a good physical appearance or in choosing the right beauty products, this will encourage her to want to watch more of their beauty product reviews and interact with them through comments on their YouTube channel. In this same vein, McCroskey et al., (1975) points out that homophily is the antecedent to PSI because such similarity between both parties boosts the development of the imaginary relationship. This idea is supported by Lee and Watkins (2016), who report that homophily is a factor that influences PSI in social media.

Like television performers, beauty vloggers are also required to show an attractive appearance while making beauty product reviews because it will persuade more subscribers to watch such contents. Hence, physical appearance becomes a primary matter for beauty vloggers, since an attractive appearance will make their subscribers be more interested in watching and paying attention to what they are explaining. As a result, subscribers will feel as if they had a connection with them and consider them as friends (Lee & Watkins, 2016). This will, in turn, encourage them to follow everything beauty vloggers share. This idea is in line with the findings of Schmid and Klimmt (2011), who affirm that physical attractiveness influences PSI: beauty vloggers’ appeal becomes a means for the development of a relationship between beauty vloggers and subscribers even though this is only a one-sided relationship driven by the subscribers’ admiration towards the physical appearance of beauty vloggers.

Social media users prefer to follow celebrities or actors whom they consider are compatible with them. An example of this is a subscriber who follows a beauty vlogger because she feels that keeping a good physical appearance is something they have in common. The fact that subscribers and beauty vloggers have similar hobbies, lifestyles, preferences, or social status (Eyal & Rubin, 2003) will encourage interaction between them. The more the similarities between them in terms of keeping a good physical appearance or in choosing the right beauty product, the higher the likelihood that they will interact with each other. In this regard, Hoffner and Buchanan (2005) state that homophily may describe subscribers’ identification with a persona due to their similar personality traits. Such a relationship may, indeed, reduce uncertainty and provide positive experiences between both parties. Therefore, homophily has an impact on PSI (Schmid & Klimmt, 2011; Sokolova & Kefi, 2020).

Beauty vloggers are expected to have a personality or attitude that is acceptable to many people, especially their subscribers, in order to increase viewers' fondness towards them (Sokolova & Kefi, 2020). Beauty vloggers may, thus, turn into a friend or partner for subscribers to obtain information about beauty products. This would certainly boost the development of a relationship between subscribers and beauty vloggers although the latter do not experience such connection directly. According to Schmid and Klimmt (2011) and Lee and Watkins (2016), social attraction is the main factor that affects PSI.

Kim and Song (2016) report that subscribers are likely to build a closer relationship with beauty vloggers on social media if they are continuously following their product reviews. They would find in beauty vloggers a reliable source of information related to beauty and cosmetics. Therefore, a good connection between both parties—based on subscribers’ trust in the contents provided by beauty vloggers—is essential, as it would encourage subscribers to purchase the beauty products reviewed by beauty vloggers (Kim et al., 2015). In this respect, Hwang and Zhang (2018) and Sokolova and Kefi (2020) point out that parasocial interaction is a factor that positively influences purchase intention.

According to the hypotheses mentioned above, we developed a research model (shown in Figure 1) based on that proposed by Sokolova and Kefi (2020) in order to investigate the effect of PSI on consumers’ intention to purchase beauty products.

Figure 1. Research model Figure 1. Research model Figura 1. Modelo de investigación Source: Taken from the study by Sokolova and Kefi (2020).

We collected data through a self-administered questionnaire and had a total sample of 151 respondents. To obtain more in-depth results, we conducted an initial screening using several criteria and selected homogeneous respondents, namely women who care about physical appearance and use beauty products in their everyday life. In addition, such selected respondents are subscribed to the YouTube channel of a beauty vlogger and have watched her beauty product reviews and interacted through comments but have never used the beauty reviewed products being.

We employed judgmental sampling, which is a nonprobability sampling technique, because respondents were chosen based on in-depth criteria, it was not possible to use a sample frame in this study.

Our model includes five constructs: physical attractiveness, attitude homophily, social attractiveness, parasocial interaction, and purchase intention. To explain each construct in this research, we relied on several previous studies to develop all the questions. In addition, we built each measurement indicator based on several theories, as described in Table 1.

Table 1. Measurement indicators Table 1. Measurement indicators Tabla 1. Indicadores de medición

Construct	Number of indicators	Source
Physical attractiveness	4	Sokolova and Kefi (2020) Lee and Watkins 2016)
Attitude homophily	4	Lee and Watkins (2016)
Social attractiveness	4	Sokolova and Kefi (2020) Lee and Watkins (2016)
Parasocial interaction	5	Lee and Watkins (2016)
Purchase intention	4	Sokolova and Kefi (2020) Lee and Watkins (2016)

Source: Created by the authors.

Such five constructs were measured on a 1–7 Likert scale, where 1 indicates “strongly disagree”; and 7, “strongly agree”. Questions were presented in the form of an online self-administered questionnaire in order to reach a wider range of beauty enthusiasts in Jakarta and its neighboring cities.

We applied a confirmatory factor analysis to test the validity of the measurement indicators and calculated their Composite Reliability (CR) and Average Variance Extracted (AVE) to assess their reliability. We performed validity and reliability tests as the first stage of data analysis using structural equation modeling and conducted the structural model once the first-stage test had been passed (Hair et al., 2009).

The obtained overall Goodness-Of-Fit (GOF) indices of the structural equation model (reported in Table 4) suggest that it is an acceptable fit to the data, with a Root Mean Square Error of Approximation (RMSEA) of 0.069 (which should be below 0.8), a Parsimony Normed Fit Index (PNFI) of 0.81, and a Comparative Fit Index (CFI) of 0.97 (Hair et al., 2009).

Table 4. Goodness-of-fit indices of the overall model Table 4. Goodness-of-fit indices of the overall model Tabla 4. Índices de bondad de ajuste del modelo general

Index	Value	Cutoff Value for GOF Index
RMSEA	0.069	RMSEA < 0.08
CFI	0.97	CFI ³ 0.95
PNFI	0.81	0 £ NFI £ 1

Source: Created by the authors.

Figure 2 and Table 5 show the results of the one-tailed hypothesis test conducted in this study. According to such results, the t-value of H1 (2.17) and H2 (4.68) is above 1.645, which means that physical attractiveness (H1) and attitude homophily (H2) have an influence on PSI. Likewise, H4 has a t-value (5.0) that is above 1.645, indicating that PSI affects purchase intention. H3, however, has a t-value of only 1.26 (which is below 1.645); hence, it is proven that social attractiveness does not have an impact on PSI in this research.

Figure 2. Output structural model Figure 2. Output structural model Figura 2. Modelo estructural de salida Source: Created by the authors.

Table 5. Hypothesis testing Table 5. Hypothesis testing Tabla 5. Prueba de hipótesis

Hypothesis		Estimate	T-value	Result
H1	Physical attractiveness ® Parasocial interaction	0.2	2.17	Accepted
H2	Attitude homophily ® Parasocial interaction	0.48	4.68	Accepted
H3	Social attractiveness ® Parasocial interaction	0.13	1.26	Rejected
H4	Parasocial interaction ® Purchase intention	0.46	5.00	Accepted

Source: Created by the authors.

Our study adds to the literature on PSI on social media (especially YouTube) and in the context of beauty products. Several theories have widely discussed PSI on traditional media, which explains an audience’s admiration towards an artist or persona they watch on films or television (Schmid and Klimmt, 2011). Nevertheless, in this digital era, this situation is different because using social media has become a topmost priority for interacting with others. Thus, this research provides a more in-depth description of the factors that influence PSI on social media sites (especially YouTube) and how this latter affects consumers’ intention to purchase beauty products.

This work is in line with previous research on PSI (Horton and Wohl, 1956; Hoffner, 1996); however, the difference is that it focuses on social media. Our findings suggest that, in the context of beauty products on social media, PSI affects viewers’ purchase intention. In addition, this study supports the results reported by Lee and Watkins (2016), who found that attitude homophily and physical attractiveness are significant factors influencing the development of parasocial relationships. This is because women will trust beauty vloggers who have their same values and the desired physical appearance, primarily in terms of beauty. The closer the relationship between both parties, the greater the viewers’ intention to purchase the products reviewed by beauty vloggers.

This study can be useful for the cosmetic industry because brands will no longer be separated from their competition. Therefore, in order to be competitive, they must develop and implement different strategies, such as using social media platforms. In this digital era, many beauty product brands in Indonesia promote their products through social media sites, mainly those that can reach a wider target market. Nevertheless, this has become a challenge because of the increasing number of brands using such tools. Thus, this work is expected to provide insights to beauty product brands into how they can make the most of social media platforms (especially YouTube) by working together with the right beauty vloggers.

According to our findings, when watching beauty product reviews, viewers will find beauty vloggers with an attractive physical look to be likeable. This attractive physical appearance will, in turn, make them feel confident that the reviewed product actually provides benefits to those who use it. Physical attractiveness does not necessarily mean having a beautiful face; it can also be represented by an attractive appearance when making a review. In addition, a healthy skin and makeup skills can be signs of attractiveness in beauty vloggers. Therefore, when it comes to choosing beauty vloggers to promote their beauty products, brands must pay attention to their physical appearance, as this aspect will make subscribers to want to watch their product reviews for a longer time and more frequently. Hence, such trust between both parties will grow and favor the development of relationships. The closer the relationship, the higher the subscribers’ interest in purchasing the product that has been reviewed by beauty vloggers.

Brands must also consider beauty vloggers’ compatibility with their target market in terms of values ​​and lifestyle because previous works, as well as this study, have reported that attitude homophily influences PSI. For instance, our results indicate that if beauty vloggers’ values ​​and lifestyle are compatible with those of their viewers, this will encourage them to watch every product review they upload on YouTube. Most respondents in this study were young female adults who naturally have specific values and lifestyles. Therefore, brands targeting this same market should choose a beauty vlogger who has their same values so that every product review will be more readily accepted and encourage viewers to purchase such products in the long term.

Additionally, this research could help beauty vloggers to better build valuable relationships with their viewers, as this will increase their intention to watch their beauty product reviews and, thus, their willingness to purchase such products. For this purpose, they should always pay attention to their physical appearance when making product reviews and build a good relationship with viewers based on these latter’s values ​​or lifestyles in terms of beauty products.