Page 103 - 臺大管理論叢第32卷第1期

NTU Management Review Vol. 32 No. 1 Apr. 2022

scholars have not reached consistent conclusions yet. Liu (2006) finds that it is the
3
volume, not the valence of online comments has explanatory power for sales. Conversely,
Chintagunta, Gopinath, and Venkataraman (2010) find not the volume, but the valence
of comments has explanatory power. And interestingly, Dellarocas, Zhang, and Awad

(2007) find that both the volume and valence of comments have explanatory power.
Although these studies have established comments as significant factors and predictors of
product sales, none of them have offered concrete models that managers can adopt in their

decision-making.
It is evident that comments represent a potentially valuable tool for streamers to
monitor viewers' attitudes in real time and adapt marketing strategies accordingly. Thus,
it is critical to develop predictive models and metrics in harnessing these comments
(Chung, 2011). As predictive models are usually forward looking, inevitably, the omission

in theoretical development causes most academic work becomes irrelevant to real-world
situations (Shmueli, 2010). To bridge the gap between methodological development
and practical application (Singleton, Mclean, and Altman, 1988), and to meet the future

trend of using predictive models to satisfy practical needs (Shmueli and Koppius, 2011),
developing a predictive model through theoretical development becomes necessary
(Shmueli, 2010).
However, prior studies have not adequately acknowledged that one model cannot fit
the entire population of streamers who could have major segments in terms of personal

characteristics. For example, Wohn and Freeman (2020) find that streamers' attractiveness
and worth have explanatory power over viewers' intention to donate during live streaming.
Wan et al. (2017) find that streamers' personalization and sociability affect the donation

intention of viewers. Unfortunately, most prior livestream studies have ignored the fact of
streamer heterogeneity (Bharadwaj, Ballings, Naik, Moore, and Arat, 2022). These studies
have treated all data alike, neglecting difference in individual-level characteristics (i.e.,

available to a multitude of people and institutions via the Internet. These terms are highly relevant
(Huang and Liu, 2016). Although this research focuses on comments, some studies on WOM and
reviews are also referred to and cited in this study.
3 Volume indicates the total number of WOM messages, while valence indicates the preference of
comments (expressed as positive/negative/neutral) (You, Vadakkepatt, and Joshi, 2015).

95