Live streaming, a real-time broadcasting service enabling live interactions and engagement, has become popular with the advent of broadband and the development of allied technologies. Live streaming platforms like YouTube Live, Twitch, and Mixer are now part of a giant industry around the globe, with as many as 30 million monthly users on Mixer, or 140 million monthly users on Twitch (Durrani, 2024). Due to the success of live streaming, media scholars have started to pay attention to this emergent media, extending the concept of parasocial relationships (PSR), which were originally developed to explain perceived connectedness or friendship between a spectator and a television character (Horton & Wohl, 1956; Rubin & Perse, 1987), to live streaming.
Recently, despite its short history, scholars have paid attention to live streaming platforms, especially in the context of the business market (Sokolova & Kefi, 2020; M. Zhang et al., 2022) or video games (Cabeza-Ramírez et al., 2022), finding that real-time viewership and interactions with a live streamer build PSR with the live streamers (Lim et al., 2020). However, less is known about the mechanisms by which PSR are developed, especially in cross-national contexts.
Thus, we aim to investigate the relationships between live streaming experiences and PSR. Through built-in interactive functions within the live streaming platforms, such as chatting or donating money, users in a live streaming session may build a sense of sociability and community among viewers (Hamilton et al., 2014; Kowert & Daniel, 2021). Perceived homophily of attitudes with the live streamer may also work as another factor that could enhance PSR with live streamers by creating an attachment to media figures who seem to be similar to their audience (Bishop, 2009; Tukachinsky et al., 2020; Turner, 1993).
This paper explores whether live streaming experience and homophily of attitudes affect PSR with live streamers, extending the previous research (see Lee et al., 2024). Both the U.S. and South Korean live streaming industries have expanded substantially over the past few years. Twitch, a U.S.-based company, has dominated the live streaming market in the United States in recent years (Durrani, 2024). In contrast, South Korea has other live streaming platforms such as AfreecaTV and Chzzk. Notably, Chzzk has grown significantly within the last year, after Twitch officially withdrew from the South Korean market (Hwang, 2024). Despite industry differences, this paper assumes that the relationships between variables would not differ significantly, as PSR has been documented as a global phenomenon (see Lim et al., 2020).
Literature Review
Parasocial Interactions and Parasocial Relationships
The concept of PSR was developed in the 1950s with the popularization of mass media at that time – radio, television, and movies (Bond, 2021). Horton & Wohl (1956) introduced the concept and defined PSR as the “seeming face-to-face relationship between spectator and performer” (p.215). Rubin and Perse (1987) explained that PSR consists of three different involvements: (a) Affectively, people will generate a sense of friendship with the media personality, (b) cognitively, people will think and understand the information during and even after the exposure to the content, and (c) people may have future behavioral involvement, such as talking about the content with friends.
The original definition of PSR by Horton and Wohl (1956) is somewhat ambiguous since it was used to refer to perceived intimacy specifically during exposure (Horton & Strauss, 1957) and it could not account for the long-term relationship built over time by repeated exposure (Horton & Wohl, 1956). Subsequent scholarship also considers parasocial interactions (PSI) (Rubin et al., 1985) to distinguish the affective components that occur during viewing from the more permanent bond that develops over time (Dibble et al., 2016). The current study follows Dibble et al. (2016) and focuses on PSR as the critical concept in the relationship between the spectator and the performer.
With the development of new media, more platforms offer different venues for people to consume content and PSR to develop in new media systems. Libers and Scharmm’s (2019) meta-analysis found that the presence of PSR has been documented by a stream of literature that includes traditional media such as radio, television, movies, and print, to emergent media that is even more interactive. Recent studies have focused increasingly on social media platforms. For example, celebrities’ self-disclosure on X and engagement in retweeting functions further enhance PSR with social media celebrities (Stein et al., 2022). Users also develop PSR with celebrities on social media such as Facebook (Wellman, 2021) and YouTube (Sokolova & Kefi, 2020). Extending prior literature on PSR and media figures in the contemporary media environment, the current study focuses on another special form of media: live streaming.
Live Streaming Services as a New Communication Platform
Live streaming is a highly interactive online service, which combines the characteristics of both traditional live TV and social media. One of the first platforms, Livestream.com, was founded in 2007. Since 2011, Twitch, a platform initially developed for gamers, has become the world’s leading live streaming platform (Durrani, 2024). Social media platforms, including Facebook and YouTube, started to offer live streaming services as well. Live streaming has become increasingly popular around the world in recent years. One of the most popular live streaming platforms in the United States, Twitch has over 17 million viewers per month. In South Korea, AfreecaTV has 2.48 million monthly users, and Chzzk has 2.27 million monthly users (Hwang, 2024). In 2023, 68% of South Koreans reported that they regularly watch live streaming platforms, and 20% broadcast live streams themselves (Statista, 2024). With a variety of content including video games, news, talk shows, cooking or eating, sports, as well as real-time communication, live streaming attracts more and more users globally (Heo et al., 2020).
Live streaming has several characteristics that differentiate it from other types of media. Firstly, unlike previous streaming services, such as television and Netflix, live streaming is not pre-recorded; it offers real-time human interactions between the streamer and viewers, facilitating their ability to communicate. In addition, interactions in live streaming platforms are transparent and reciprocal. Although some radio and television programs also offer opportunities to provide feedback, the number of comments sent, what is said, and what gets curated on the feed remains opaque. In contrast, one of the common affordances in live streaming platforms is to present all comments on the screen, helping to imitate a face-to-face conversation and increase viewer-streamer interactions. Second, the live streaming session contributes to constructing a virtual community between the audience and the streamer. Not only can the live streamer and the audience interact with each other, but the audience can also communicate using the live commenting windows, adding another possible interaction that can be considered a PSR. This kind of function fulfills people’s social-based motivation and creates a sense of community (Hilvert-Bruce et al., 2018).
Parasocial Relationships with Live Streamers: Live Streaming Engagement and Homophily of Attitudes
Then what factors could enhance one’s PSR with live streamers? First, frequent exposure to media has been considered to foster PSR (see Levy, 1979; Rubin et al., 1985). People who consume a lot of media, regardless of content and format, are more likely to form PSR with media personalities (Schiappa et al., 2007). There are other functions on live stream platforms for people to engage with. As mentioned in the previous section, audiences can express their opinions live and receive reactions from the streamer or other members of the audience. Moreover, many live streaming platforms also offer donation functions to financially support a streamer. It is widely accepted that perceived responsiveness promotes PSR with media figures (Rubin & Step, 2000; Tukachinsky & Stever, 2019). Aligning with this, Hilvert-Bruce et al. (2018) conclude that there are at least four social motivations of live stream engagement: emotional connectedness, time spent, subscription, and donation. Rihl and Wegener (2019) introduce a measurement for PSR with YouTubers, including commenting on content and favorite YouTubers commenting back to users. Based on these and given the characteristics of live stream platforms, we hypothesize that:
H1a: Frequent exposure to live streaming is positively related to parasocial relationships with live streamers.
H1b: The number of channels a user is subscribed to is positively related to parasocial relationships with live streamers.
H1c: Expressing opinions during live streaming is positively related to parasocial relationships with live streamers.
H1d: Receiving reactions from live streamers is positively related to parasocial relationships with live streamers.
H1e: Donating money to streamers is positively related to parasocial relationships with live streamers.
Homophily is the tendency to connect and interact with people who are similar to oneself (McPherson et al., 2001). The characteristics and qualities people share vary by age, gender, race, ethnicity, education, class, political identity, and attitudes. This homophily principle may result in negative social consequences, such as the clustering of like-minded people that might increase social and political polarization (Bishop, 2009), but despite the normative appeals of diversity, there is evidence that given the freedom to choose, people interact with content and people who share certain characteristics with them. In the PSR literature, one of the types of user-figure interactions is similarity identification, a mechanism by which audiences find similarities and share similar perspectives (Giles, 2002; Rosengren & Windahl, 1972).
We were particularly interested in the perceived homophily of attitudes, that is perceiving that others share one’s attitudes, which can be regarded as a strong predictor of the development of PSR. Homophily of attitudes can reinforce relationships (Rubin & Step, 2000) and can be more important than demographics or appearance when developing PSR with media characters (Tukachinsky & Stever, 2019). Empirically, many studies confirm that homophily of attitudes is the strongest predictor of PSR development in a variety of contexts, such as with fictional characters, soap opera characters, newscasters, and comedians (Schmid & Klimmt, 2011; Turner, 1993). Recently, Sokolova & Kefi (2020) also found that participants’ homophily of attitudes with influencers on social media is positively related to their PSR, which further influences the audience’s purchase intentions. Thus, this study hypothesizes that:
H2: Perceived homophily of attitudes is positively related to parasocial relationships with live streamers.
Method
Data
Data were collected using online survey panels administered by Qualtrics. The sample was designed to represent the general population of adults in the United States and South Korea using quota sampling. The sampling process of Qualtrics involved two stages. In the first stage, subjects were randomly selected from the online panel in each country created based on demographic and geographic information. In the second stage, subjects’ profiles were generated by balancing demographics. Then Qualtrics gave access to the survey to their pool of respondents, through a web portal or an email invitation. To avoid self-selection bias, the survey did not explain study details but explicated that the survey was only for research purposes. Before respondents began the survey, they viewed a consent form that clarified the purpose and procedures of the survey, a privacy notice, and contact information. Both surveys completed IRB procedures at the university.
The U.S. survey was completed between February and March 2020 with 510 respondents. The sample reflects well the U.S. adult population in terms of gender (sample: 49.2% males; census[1] 49.2% males), race (sample: 76.6% White; census: 76.3% White), income (sample median: between $3,001 and $4,000/month; census: $3936) and educational level (sample: less than high school 9.2%, high school 36.3%, incomplete college 24.1%, college graduates 18.0%, more than college 12.4%; census: less than high school 10.0%, high school graduates 28.8%, college no degree or associate degree 26.0%, bachelor’s degree 22.5%, more than bachelor’s degree 13.4%), but is older than the general population (sample median = 46.5; census median = 37.2).
The South Korean data was collected between March and May 2020 and includes 504 completed responses. The sample reflects the general population well in terms of gender (sample: 50% males; census[2] 49.6% males). The sample is slightly wealthier than the population when considering monthly income (sample median: between 3 million won and 5 million won, which is approximately $2,200-$3660; census: 2.2 million won or $1,610), more educated than the population (sample: less than elementary school 2.2%, middle school 6%, high school 28.8%, more than high school but no bachelor’s degree 3.4%, bachelor’s degree or community college degree 61.5% and more than graduate school 0.4%; census: less than elementary school 11.8%, middle school 10.4%, high school 38.6%, community college and bachelor’s degree or more 39.3%) and somewhat older than the population (sample median= 48; census median = 43.1). The race item was not included in the Korean survey since 95.4% of the population shares the same ethnicity in the country (Korean Statistical Information Service, 2019).
As our study pertains to people who use live streaming services, measures and analyses are based on a subset of the sample who fit this category. Compared to the full sample the subset that uses live streaming services is: 77.8% in South Korea (n = 392) and 65.9% in United States (n = 336) watch live streams.
Measures
Table 1 provides summary statistics of variables described below and Appendix A illustrates the question wording employed for all items.
Live Streaming Use
First, we asked about the frequency of live streaming usage providing respondents with a clear definition of live streaming: “Live streaming is defined as real-time video broadcast via social media platforms, like Facebook, YouTube, Instagram, or Twitch. How frequently do you watch live streaming?”. This question employed a scale from 0 = never to 5 = frequently and was used to screen out non-live streaming users in the sample.
Then, using a similar scale, we asked how frequently respondents engaged in the following activities: (1) donating money and (2) leaving comments or chatting in the live stream chat room. To measure interactivity with the streamer, the survey asked, “When you have commented to the streamer, how frequently do they react back to your comments?”, employing a similar 0 = never to 5 = frequently scale. The study also inquired about the number of channels the subject followed or subscribed to.
Parasocial Relationship
There are several scales developed to evaluate people’s PSR. Dibble et al. (2016) found that the classic PSI scale by Rubin et al. (1985) corresponds well with long-term liking and bond with media characters. Since our research focuses on the long-term relationship rather than one-time exposure, we adapt the items from the PSI index developed by Rubin et al. (1985), originally designed for TV newscasters.
To adapt Rubin et al.'s (1985) scale, the research team conducted a pretest survey (n = 239) at a large midwestern university in the United States, using twenty items from Rubin et al. (1985) and changing “a newscaster” to “a live streamer”. The battery of items in the pretest was highly reliable (α= .93). The top 5 items with robust inter-items correlations (α = .91) were selected for use in the main study to measure PSR with live streamers.
In order to gauge PSR with one streamer, the survey initially asked who the favorite streamer/channel was. This follows previous designs to gauge PSR with one streamer, not PSR with multiple media figures (Rihl & Wegener, 2019). The questions employed were: (1) When I’m watching a live stream, I feel as if I am part of their group, (2) The streamers make me feel comfortable, as if I am with friends, (3) I like hearing the voice of my favorite streamer at home, (4) My favorite streamers keep me company when the streaming is on, and (5) I look forward to watching my favorite streamer when he or she is on streaming.
Homophily of Attitudes
An index was created based on McCroskey et al. (1975)'s work. Respondents were asked to report perceived homophily of attitudes by answering two questions that were placed at opposing ends of a continuum of bipolar statements asking (1) if their favorite streamers think like them or don’t think like them and (2) whether they are similar to the respondent or different from them.
Demographics
In the analyses, certain demographic variables are controlled for: age, gender, education, and income in both countries. Education was measured with a scale ranging from 1 to 7 (1 = None; 7 = Graduate) and income was measured with an 8-point scale (1 = less than $1000 [less than 1 million won]; 7 = more than $20,000 [more than 20 million won]). For the analysis of the U.S. dataset, we included a race variable which was recoded as a binary (0 = Non-White; 1 = White).
Results
Table 1 illustrates summary statistics of variables. In the regression analyses, two items of homophily of attitudes were combined as their correlations were high both in Korea (r = 0.63, p < .001; M = 3.77, SD = 1.15) and the United States (r = 0.66, p < .001; M = 4.14, SD = 1.29). The dependent variable, PSR with live streamers measured with 5 items was collapsed into a mean index and the index was highly reliable in both Korea (α = 0.91, M = 2.57, SD = 1.15) and in the United States (α = 0.95, M = 2.87, SD = 1.46).
To test H1a to H1e and H2, parallel hierarchical regressions predicting PSR in each country were run (see Table 2). Model 2 uses live stream usage as the main predictor and the model fits well in both South Korea (R2 = .37, F (9, 381) = 26.12, p < .001) and the United States (R2 = .41, F (10, 325) = 23.95, p < .001). R2 increased significantly when comparing Model 1 and Model 2 in both countries, indicating that live streaming usage in both countries was a strong predictor of PSR with live streamers.
In South Korea, all the variables related to live stream use were positively related to PSR. As expected, the frequency of watching live streams (B = .19, p <.001) is positively related to increased PSR with a streamer. The number of channels they subscribe to or follow also matters (B = .004, p = .009). Donating money (B = .17, p = .001), expressing opinions in the live chat room (B = .17, p <.001), and receiving reactions from a streamer (B = .09, p = .026) promote PSR.
A similar pattern was found in the United States. Exposure to live streams is the strongest predictor (B = .21, p <.001), but the number of channels does not have a significant relationship with PSR (B = .001, p = .361. Engaging in activities like donating money (B = .12, p = .009), commenting in the chat (B = .17, p = .002), and receiving reactions from a streamer (B = .13, p = .011) are associated with PSR. Overall, PSR with a live streamer is enhanced by activities and experiences while watching live streams as well as exposure to live streams. Thus, H1a to H1d are generally supported, but H1b is only supported in South Korea.
To test H2, another block is added in Model 3. The explanatory power of the model improves both in South Korea (R2 = .45, F (10, 380) = 32.28, p < .001) and the United States (R2 = .48, F (11, 324) = 28.56, p < .001). Homophily of attitudes has a strong positive relationship with PSR with live streamers in South Korea (B = .31, p < .001) and in the United States (B = .34, p <.001), and most of the variables related to live stream use remain significant. As postulated in H2, if viewers perceive a similarity between a streamer and themselves, this perception can result in increased levels of PSR. Thus, H2 is supported.
Discussion
The current study found that both the U.S. and South Korean live stream usage, engagement, and perceived homophily of attitudes are positively associated with experiencing a PSR with a live streamer. The positive relationship between live streaming engagement and PSR improves our understanding of the development of PSR in new online platforms. Methodologically, the current study updates the original PSR scale created by Rubin et al. (1985) for live streaming, incorporating suggestions of prior research (see Lim et al., 2020).
Engagement during live streaming had a positive relationship with PSR with the live streamer. This aligns with research on PSR, which suggests that more frequent viewership leads to a more positive attachment (Rubin et al., 1985; Schiappa et al., 2007). Other types of engagement, such as donating money and expressing opinions in chat rooms, also fostered PSR. This indicates that active engagement and the consumption of financial and cognitive resources promote PSR with live streamers (W. Zhang, 2022). Moreover, the live streamer’s responsiveness can foster PSR, as they play the role of community leaders (Stein et al., 2022).
Furthermore, the findings reveal significant similarities between South Korea and the United States. The measurement of live streaming use, homophily of attitudes, and PSR did not differ between the two countries, in terms of reliability and their relationships with the dependent variable. In addition, homophily of attitudes has important effects on PSR, showing that viewers tend to feel more intimacy with streamers who “seem to have similar attitudes” to them. This corresponds to the meta-analysis conducted by Tukanchinsky et al. (2020), showing that multiple studies find homophily can facilitate PSR with media figures in diverse settings. This may be due in part to the global nature of PSR with live streamers (see Schmid & Klimmt, 2011). While the live streaming environments in both countries were similar, they also showed some differences. For example, there was no significant relationship between the number of channels followed/subscribed to and PSR in the United States, whereas this relationship was significant in South Korea.
Despite the contributions of this study, several limitations suggest the need for future research. First, the measurement of homophily of attitudes was measured with two items that do not consider all possible dimensions of homophily. Other types of homophily, including appearance, value, and background homophily, could be associated with the formation of PSR with live streamers (McCroskey et al., 1975). Though the correlation between the two items employed was high, and a significant relationship between homophily of attitudes and PSR was found, future research that considers additional items would be beneficial. In addition, the question of whether an individual forms a PSR with only one favorite streamer or with multiple streamers at the same time is also important. Future studies should address this issue.
In addition, the question of the causal relationship remains. Many scholars suggest that media use and engagement can promote PSR, but they also do not rule out a feedback loop from media consumption to PSR. Enhanced PSR could lead to media use such as re-watching shows or seeking further information (Tukachinsky & Stever, 2019), or in our case consuming more live streams from the preferred live streamer. This same cyclical logic could be argued for the engagement measures employed and PSR. While this study has followed previous theorizations in setting up the models, the cross-sectional nature of our design does not allow for a definitive test of causality among the elements in the model. Future studies using longitudinal data should consider these potential “virtuous” circles investigating the possible linkages over time.
Conclusion
Overall, despite the limitations, the current research contributes to a better understanding of the development of PSR in live streaming platforms and theoretically provides insight to connect classic PSR theory in new media platforms and identify factors affecting PSR. The result regarding the relationship between homophily of attitudes and PSR links to a recent surge of research on perceived polarization (Armaly & Enders, 2021; Banks et al., 2021; Yang et al., 2016), as it denotes a perceived similarity or distance from others. This finding calls for investigating live streaming sessions as potentially homogeneous online communities with shared thoughts and ideas that can further exacerbate bigotry and social enclaves.
The study also demonstrates that similar results were found in South Korea and the United States. We call for attention to extensive and exhaustive research on the mechanism of PSR and its societal implications in multiple countries from diverse cultural backgrounds. In this study, despite important cultural differences between South Korea and the United States, the results are mostly convergent in terms of some of the antecedents of PSR. However, future studies that focus more on the consequences of PSR might reveal different patterns of how PSR plays out in different cultural milieus. For example, would the consequences of PSR be stronger in countries that are more collectivistic or individualistic?
In addition, this study has contributed by providing insights into how different types of engagement—such as donations and chat interactions—can be leveraged to build stronger PSRs. Recognizing the role of factors in developing PSR can help live streamers understand how to connect with their audiences, promote positive community interactions, and enhance viewer loyalty and satisfaction.
In conclusion, as the popularity of live streaming continues to grow globally, future studies can consider broadening their perspectives to analyze more countries and narrowing down the context into specific content such as politics or video games. Future research should explore and unpack the dynamics of PSR across different cultural contexts to better understand and address its implications for users around the world.
Funding
The project was funded by the University of Wisconsin Foundation.