More than the Quantity: The Value of Editorial Reviews for a UGC Platform
Management Science 68(9), pp 6355-7064.
48 Pages Posted: 16 Oct 2019 Last revised: 4 Nov 2022
Date Written: July 29, 2021
Abstract
We investigate an editorial review program for which a review platform supplements user reviews with editorial ones written by professional writers. Specifically, we examine whether and how editorial reviews influence subsequent user reviews (reviews written by non-editor reviewers). A quasi-experiment conducted on a leading review platform in Asia, based on several econometric and natural language processing techniques, yields empirical evidence of an overall positive effect of editorial reviews on subsequent user reviews from the platform’s perspective. More reviews are provided for restaurants that receive editorial reviews, suggesting a herding effect, as users are more inclined to follow editors’ choice of restaurants to review. These reviews discuss substantive topics while also including a discussion on other topics, leading to a net increase in content length and variety. They also are more neutral in sentiment and are associated with lower rating valences. Further analysis of the mechanism reveals that the subsequent user reviews of the restaurants that receive editorial reviews become more similar to the editorial reviews in regard to topics, sentiment/rating, length, and readability, indicating a herding effect in how to write a review as the main driver of the change in the subsequent reviews. We further empirically isolate this herding effect among long-time reviewers. The findings suggest that review platforms could use an editorial review program to not only boost review quantity but also manage the content quality. By supplementing high-quality editorial reviews with user reviews, the platform can improve the overall content quality of user reviews.
Keywords: natural language processing, editorial review, user-generated content, difference-in-differences, herding effect
JEL Classification: M00
Suggested Citation: Suggested Citation