Research On The Existence And Depending Factors Of Third-party Online Reviews' Bias And Sequential Bias

Online reviews on propagation characteristics and information transmission characteristics are different from the traditional transmission of public praise. So it has been a focus of market researchers since its birth. Third-party site is an independent review platform, although most people think that the third-party site is relatively independent and credible, and most commentators are also relatively independent and unbiased. A research confirmed that there are different types of bias in online reviews. This research based on the most mainstream third-party review platform Douban, using time series model and panel regression model to prove the existence of review bias and sequential bias which depends on the dimensions of reviewers, content of comments and rating characteristics.Firstof all, let's determine that whether there is bias in third-party online reviews. W e collected 3657 reviews of 11 movies in this research within five weeks after the movie s hows to assess the bias of third-party reviews by using one of the classic model Hodrick-P rescott filter, finally we found out that each film score have different degrees of bias. At t he same time, we use the ARMA model and impulse response function to analyse the cha racteristics of score bias, and found out that score bias shows varying degrees of autocorre lation, it also possesses a certain amount of time variation characteristics, and the gap of s core bias can also be showed by article length, extreme score, comments usefulness votes and reviews attention degree. Then we use panel data model to analyse the correlation of stage t score and stage t-1 variables, which confirmed that the existence of the third-party reviews bias, and we also found out that the sequential bias of stage t depends on the articl e length,the number of movies that commentators have seen and extreme score of stage t-1. Finally, draw conclusions depending on those research, and put forward suggestions to management of consumers, businesses and websites, and point out the insufficiencies of t his study and future research direction.