Park, Jaecheol, Joy Wu, Arslan Aziz, Gene Moo Lee. “Platform Intervention and Ecosystem Response: Evidence from Amazon’s Ban on Incentivized Reviews”, Working Paper.
- Presentations: UBC (2021), KrAIS (2021), WISE (2021), PACIS (2022), SCECR (2022), BU Platform (2022), CIST (2022), BIGS (2022)
- Preliminary version in PACIS 2022 Proceedings
- RAs: Minsuk Seo, Vibudh Singh
Online marketplaces depend on consumers to voluntarily produce the review information that other consumers use to make purchase decisions. To accelerate this production, many platforms have allowed sellers to provide free or discounted products in exchange for reviews that disclose the incentive. How such seller-subsidized content affects the supply of organic reviews, and how sellers respond when a platform withdraws this option, remain unclear. We theorize two competing channels through which such reviews shape organic contributions: an information crowd-out effect, when readers treat incentivized content as part of the stock of product information, and a skepticism effect, when readers process the disclosure as a signal of seller persuasion intent. Exploiting Amazon.com’s ban on disclosed incentivized reviews as a natural experiment, we identify the causal effects of the ban on organic reviews. Using a fine-tuned language model to detect incentive disclosures in 15.3 million reviews and difference-in-differences estimation with matched control products, we find that the ban increased the weekly number of organic reviews by 5% and their length by 9%, and increased images and helpful votes, while leaving rating and sentiment unchanged, consistent with information crowd-out. Further analyses of post-ban seller strategies show that these findings are not driven by sellers’ shift toward undisclosed incentivized reviews. We also show that review-reader attention moderates the relative strength of the two mechanisms: the information crowd-out (skepticism) effect is more pronounced when attention level is low (high). An online randomized experiment confirms the psychological underpinnings of our theory. For platform operations, the findings show that incentive programs used to address the cold-start problem carry a hidden cost in displaced organic information, and that review policies should be evaluated by the supply of organic content and by sellers’ compliance responses, not by ratings alone.