Vertical Integration and Consumer Choice: Evidence from a Field Experiment

Platforms, retailers, and other firms often offer their own products alongside products sold by competitors. We study the effects of this practice by combining a field experiment that hides brands owned by Amazon (i.e., private labels) from shoppers on Amazon.com with model-based counterfactuals and welfare analysis. In the absence of private labels, consumers substitute toward products that are similar along most observable dimensions. Removing Amazon brands does not change consumers' search effort or their propensity to shop at other retail websites. Despite the ample availability of observably similar alternatives, our welfare estimates imply that, for the categories we study, removing Amazon brands would reduce consumer surplus by 5.5 percent in the short run, with approximately 10 percent of the impact due to equilibrium price increases by other sellers. The effects are heterogeneous, with consumer surplus reductions exceeding 10 percent in some categories, while other categories realize much smaller decreases when Amazon brands are removed. Demoting private labels in search results to counteract potential self-preferencing does not lead to gains in consumer surplus. This outcome arises because a subset of consumers derive greater utility from private labels and benefit from their high placement in search results.