Page 104 - 臺大管理論叢第32卷第1期

Internet Celebrity Economy: Exploring the Value of Viewers’ Comment Features and Live Streamers’
Marketing Strategies in Forecasting Revenue

streamer heterogeneity, which creates challenges to data analysis. And when these analyses
are hard to execute rigorously, other researchers are not able to replicate it (Allenby, Rossi,
and McCulloch, 2005). Thus, considering the individual heterogeneity in the predictive
model is necessary (Lee, Cho, Lee, and Lee, 2006; Van den Bulte and Joshi, 2007). To

fill this research gap, this study applies the Hierarchical Bayesian (HB) model to handle
sample heterogeneity.
This study provides an overarching framework that addresses the existing deficiencies

aforementioned by exploring the predictive value of viewers' emotion-embedded
comments, comment metrics, and streamer heterogeneity and behavior on streamers'
revenue from live streaming. In other words, we propose a revenue forecasting model
tailored to the livestream industry, and test streamers' performances in the context of live-
streaming revenue forecasting. Surprisingly, we find that prior livestream studies that

examined the factors influencing streamers' revenue may have offered biased conclusions
because they ignore the impacts of streamer heterogeneity.

2. Literature Review and Inference Developement

This section introduces live-streaming platforms and proposes three theories to
explain the variables that predict viewers' gift-sending behavior on a live-streaming
platform presented in prior literature.

2.1 Live-Streaming Platforms
In recent years, live-streaming platforms have emerged as a unique form of social
media (Hu, Zhang, and Wang, 2017). With the help of cutting-edge internet technologies,

live-streaming platforms offer text and image features and audio and video functions.
Unlike traditional social media, such as virtual communities and blogs, a live streaming
platform allows its users (e.g., internet celebrities; also known as wanghong in Chinese) to
stream performances, games, or daily life activities on a real-time basis. These streamers
can also engage in dialogue or interact with their viewers while streaming. Viewers, in

turn, can not only chat with each other through text-based messages but also leave real-
time comments for the streamers, which appear in the form of a ticker across the screen,
and send virtual gifts to the streamers (see Appendix for the gift examples). The virtual

96