Optimal Rating Systems on Platforms

Résumé

This paper considers the design of an optimal rating system, in a market with adverse selection. We address two critical questions about rating design: First, given a number of categories, what are the criteria for setting the boundaries between them? Second, what are the gains from increasing the number of categories? A rating system helps reallocate sales from lower- to higher-quality producers, thus mitigating the problem of adverse selection. We focus on two main sources of market heterogeneity that determine the extent and e↵ect of this reallocation: the distribution of firm qualities and the responsiveness of sellers’ supply to prices. We provide a simple characterization for the optimal rating system as the solution to a standard k-means clustering problem, and discuss its connection to supply elasticity and the skewness of firm qualities. Our results show that a simple two-tier rating can achieve a large share of full information surplus. Additionally, we characterize the conflicting interests of consumers and producers in the design of a rating system.