A Theory of Likes

Abstract

We study binary, non-strategic ratings when consumer preferences reflect a combination of objective quality differences between products and idiosyncratic taste. A rating then decomposes into objective and subjective content. Rating systems that are extremely strict or extremely lenient only convey objective content, maximizing consumer welfare when the population is not biased toward a particular product. With a biased population, less extreme rating systems that convey subjective content can be valuable. In many environments, a sales-maximizing platform prefers lenient rating systems, creating a conflict with welfare maximization.