USC Marshall School of Business, University Of Southern California, United States
E-commerce platforms use recommendation algorithms to select products and services for consumers from a vast number of options. However, analyzing the causal impact of recommendation algorithms is challenging due to unobserved selection processes, hidden variables, and endogenous adjustments to the current states of the marketplace. We leverage Amazon’s binding commitment to comply with the U.K. Competition and Markets Authority’s (CMA’s) regulation, which prohibits Amazon from using Prime eligibility as a criterion for recommending the Featured Offer within a product market, to study the causal impacts of the recommendation algorithms on seller niche market entry, price and competition strategies, consumer search and purchase behaviors, and the revenues of sellers and the platform. We first identify that the binding commitment to the CMA causes the platform to shift recommendation weights from Prime eligibility to low offer price. Then, we study the average treatment effects of the recommendation algorithm change on product market outcomes and further analyze driving mechanisms. We find that Amazon’s recommendation algorithm change in response to the CMA’s regulation increases product total clicks while decreasing product ratings and consumer click-through rates. Mechanism analysis shows that the recommendation algorithm change decreases offer prices while decreasing offer qualities (intensive margin) and offer varieties (extensive margin), causing sellers to race to the bottom and decreasing consumer satisfaction. Our causal analysis of the impacts of the platform recommendation algorithm provides insights for third-party sellers responding to the platform algorithm change, for the platform owners to design the recommendation algorithm, and for the regulators to design policies in guiding the recommendation algorithm design to promote competition fairness and enhance consumer welfare.
