Predicting the future increment of review helpfulness: an empirical study based on a two-wave data set

Abstract

Purpose Identifying and predicting the most helpful reviews has been a focal interest in the fields including information management, e-commerce 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 Yelp.com 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.

Factors correlated with the perceived usefulness of online reviews for consumers: a meta-analysis of the moderating effects of product type

“…(2) Review replies, review depth, review pictures, positive reviews and review time have been highlighted over the past decade as common review-level factors that correlate with review usefulness (Pan et al, 2021). The meta-analysis in this study also confirmed a significant relationship between the above aforementioned factors and perceived usefulness, but the degree of correlation differed.…”

“…(1) Existing studies have mainly focused on the impact of reviewer characteristics and review contents on review usefulness. Insufficient attention has been paid to the impact of the review reader's characteristics (Pan et al, 2021). People with a high tendency to trust are more willing to trust strangers, and the existing literature also verifies that whether consumers are willing to trust online reviews depends on trust tendency (Filieri, 2015;Luo and Li, 2013).…”

“…(3) Reviewer characteristics are another group of factors that have a correlation with the review helpfulness. The reviewer information disclosure, expertise and trustworthiness, which are the most discussed features, have positive correlations with perceived helpfulness of review (Liu and Park, 2015;Pan et al, 2021). The recipients and disseminators of online reviews are mostly strangers.…”

Consumers with specialised and diverse experience produce more helpful reviews

“…As shown in Figure 1, we regard the round 12 data as the historical records, from which the reviewer features are extracted, including especially the experience distribution. Inspired by the framework proposed by Pan et al (2021), the reviews posted between t1 and 14 September 2018 (t′) are the target reviews for the latter econometric analysis. The helpfulness of the target reviews, i.e.…”