Identification and Impact of Online Deceptive Counterfeit Products : Evidence from Amazon

With the proliferation of third-party sellers, counterfeiting has become a serious source of friction in online marketplaces. We apply natural language processing techniques to Amazon product reviews to generate a listing-level counterfeit probability, which in turn is used to classify ASIN (Amazon Standard Identification Number) listings as likely counterfeit or a likely authentic. We focus on two product categories, one a taste-based experience good (men’s fragrances) and the other being a utilitarian product (wireless cell phone chargers). We embed the estimated counterfeit probability into a BLP-type choice model to investigate how counterfeiting intensity affects user demand and platform revenues. We confirm that consumer disutility is increasing in the counterfeit probability, more so for high-end or popular products. We further find a substitution effect between likely counterfeit and likely authentic products: a 10% decrease in the price of a likely counterfeit product is associated with an average 0.0011% decrease in the market share of a likely authentic product. We leverage our structural parameter estimates to run a number of counterfactual experiments. These experiments suggest that protecting authentic sellers by simply banning all likely counterfeit listings would drastically reduce platform revenues. Instead, the deployment of counterfeit detection algorithms, and reporting the results to users, would align the interests of authentic sellers with the welfare of the platform. Overall, our analysis provides a robust empirical platform for identifying deceptive online counterfeiting, and understanding its impact on the various stakeholders of an online retail platform