영화확산에 대한 온라인구전의 차원 및 영향력 분석

In this study, using Korean movie box office data and online review data, we explore the 4 dimensions of eWOM(electronic(or online) word-of-mouth) (volume, valence, variance, and sentiment), and their effects on the number of audiences according to the analysis method. The volume, valence, and variance of eWOM were measured by the number of reviews, average rating, and rating variance, respectively, while the sentiment of eWOM was measured by the average sentiment scores derived by the BERT(Bidirectional Encoder Representations from Transformers) model. As a result, it was recognized that we have to be careful in interpreting the analysis results using secondary data, because the units of measurement of the 4 dimensions(the number of reviews, average rating, rating variance, and average sentiment) of eWOM are not all same, and the 4 dimensions are highly correlated with each other. By comparing the estimation results of three analysis models: regression model for aggregate data, panel regression model and panel error correction model for weekly panel data, it was found that the effects of 4 dimensions of eWOM may differ depending on the analysis method. The analysis result from the panel error correction model showed that only the volume of eWOM(number of reviews) had a significant effect on the number of audiences. The results of this study empirically show that considering the characteristics of the data used and the selection of an appropriate research method are important for accurate analysis, especially when we are trying to analyze the effects of the dimensions of eWOM on the movie box office success using secondary data about online movie reviews.