The Extreme Distribution of Online Reviews: Prevalence, Drivers and Implications
68 Pages Posted: 15 Jan 2018
Date Written: January 11, 2018
Online reviews are often criticized for being uninformative due to an extremely high percentage of near-perfect ratings, making it difficult for consumers to distinguish between high and low quality products. In this research, we investigate the prevalence and possible reasons underlying the so-called extreme distribution of reviews in which reviews are heavily skewed to the positive end of the rating scale, with a few reviews in the mid-range and some reviews at the negative end of the scale. Putting together a large dataset of online reviews - over 130 million reviews from 18 major online platforms - we find that the extreme distribution of reviews occurs on different online platforms, but the platforms vary with respect to the prevalence of the extreme distribution of their products based on how selective customers are in reviewing products on the platform. Using a multi-method approach, including secondary data, experiments as well as survey data, we identify multiple potential drivers underlying the extreme distribution. Our results demonstrate that polarity self-selection, described as the higher tendency of consumers with extreme experiences to provide a review, is the main driver of the extreme distribution. We also find evidence that cognitive dissonance further enhances the positive skewness of the rating distribution, while fake reviews in contrast enhance negative extreme ratings. Additionally, we describe and demonstrate that polarity self-selection and the extreme distribution reduce the informativeness of online reviews, thus explaining the inconsistent findings in previous research regarding the relationship between average ratings and sales, as well as the more consistent relationship between review volume and sales. Finally, we propose potential mechanisms to reduce this bias in online ratings.
Keywords: User Generated Content, Online Reviews, Online Ratings, Extreme Distribution, Self-Selection
JEL Classification: M30, M31
Suggested Citation: Suggested Citation