66 Pages Posted: 24 Apr 2017 Last revised: 6 Feb 2019
Date Written: January 28, 2019
Ratings are often coarse. Examples include corporate debt ratings, stock analyst recommendations, Morningstar ratings, bank analyst evaluations, referee suggestions to editors, student grades, Yelp ratings, and film reviews. In the classic Crawford-Sobel (1982) model, coarse messaging is not optimal if the sender's and receiver's interests coincide. In the situations listed above, senders' and receivers' interests likely do coincide and the rating systems appear to be deliberately chosen. We provide a cheap talk model in which messages are received with exogenous noise and show that the optimal rating system may be coarse: while coarse messages are less precise, they are easier to interpret. In numerical work, we derive predictions for the distribution of ratings and show that the distributions we see in practice match the predictions of the model.
Keywords: Cheap Talk, Discrete, Coarse, Yelp, Credit Ratings, Stock Analysts
JEL Classification: D83, G00, M00
Suggested Citation: Suggested Citation