Page 123 - 臺大管理論叢第32卷第1期

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

psychology, such as social interaction (Yu et al., 2018; Zhou et al., 2019), attachment
(Wan et al., 2017), emotion (Phonthanukitithaworn and Sellitto, 2017), attractiveness
(Wohn and Freeman, 2019), and reciprocity (Zhang et al., 2019). Our study is the first to
synthesize an analysis of paid gifting behavior in live streaming from the perspectives of

viewers’ various comment metrics, discrete emotion-embedded comments, and streamers’
characteristics and behaviors. This study contributes to the literature by demonstrating that
the effects of different comment metrics and discrete emotions on gift-sending behavior in

live streaming depend on streamer heterogeneity.
In addition, Zhou et al. (2019) examine viewers’ comments and gift-sending
behavior using a fixed period (one minute) as an analysis unit is not reasonable. Because
the viewers’ behaviors are affected by streamers’ behavior, and viewers’ behavior is the
result of their considerations, a viewer’s behavior will have a carryover effect. Therefore,

Zhou et al. (2019) research can not identify the distribution patterns hidden in streamer
behaviors and gift-sending behavior, possibly causing measurement errors. This study
corrects the analysis unit of time to precisely explore the viewers’ comments and gift

sending to identify effectively that the true distribution patterns of their behaviors depend
on streamers’ behaviors. Furthermore, past studies related to gift sending have only
counted the number of gifts (Zhou et al., 2019), ignoring the unequal value of each gift,
which in turn have caused measurement errors. Since DouYu offers many types of gifts,
and the value of each gift differs, this study revises the coding of the gift values according

to the gift values provided on the DouYu website.
Furthermore, most previous studies have employed self-reporting methodology (i.e.,
survey questionnaires and in-depth interviews) to survey streamer revenue (Wan et al.,

2017; Wohn and Freeman, 2020; Zhang et al., 2019). This study adopts the following two
approaches for data collection. First, it employs data crawling to examine the data on gift
sending and second-by-second chat records from DouYu. Second, we conduct content
analysis to gather information on streamers’ behaviors. Both approaches produce abundant
data on viewers and streamers and thus facilitate more accurate model results. These data

sources also strengthen the internal validity and model accuracy.
Finally, prior studies of paid gifting in live streaming have used regression (Yu et al.,
2018; Zhou et al., 2019), structural equation modeling (Phonthanukitithaworn and Sellitto,

2017; Wan et al., 2017), and content analysis (Zhang et al., 2019). This study contributes

115