Predicting The Future Increment Of Review Helpfulness: An Empirical Study Based On A Two-Wave Data Set

Xue Pan,Lei Hou,Kecheng Liu

DOI: https://doi.org/10.1108/EL-06-2020-0130

2021-01-01

Abstract:Purpose - Identifying and predicting the most helpful reviews has been a focal interest in the fields including information management, ecommerce and marketing, etc. Though many factors are found correlated to the helpfulness of reviews, they may suffer endogeneity problems, as normally the data is observed in the same time window. This paper aims to tackle such a problem by examining the predictive power of different factors on the future increment of review helpfulness.Design/methodology/approach - Adopting a longitudinal data of 443 K empirical business reviews from collected at two different time points, six groups of predictors are extracted from the first snapshot of data to predict the helpfulness increment of old and recent reviews, respectively, between the two snapshots.Findings - It is found that these factors in general are with moderate accuracy predicting the helpfulness increment. A different group of features shows quite different predictive power. The reviewer disclosure information is the most significant factor, while the review readability does not significantly improve the accuracy of prediction.Originality/value - Instead of the total number of helpful votes observed in the same time window with the explanatory variables, this paper focuses on the future increment of helpful votes observed in the following time window. With such a two-wave data set, the endogeneity problem can be avoided and the explanatory factors for review helpfulness can, thus, be further tested in the prediction scenario.