An enduring puzzle is why credit rating agencies (CRAs) use a few categories to describe credit qualities lying in a continuum, even when ratings coarseness reduces welfare. We model a cheap-talk game in which a CRA assigns positive weights to the divergent goals of issuing firms and investors. The CRA wishes to inflate ratings but prefers an unbiased rating to one whose inflation exceeds a threshold. Ratings coarseness arises in equilibrium to preclude excessive rating inflation. We show that competition among CRAs can increase ratings coarseness. We also examine the welfare implications of regulatory initiatives.