We develop a credit-risk model with dynamic credit-rating revisions. A credit rating agency that assigns a credit rating on a firm considers several factors such as reputation, business relationship, and rating stability when revising the rated entity’s current rating. The model exhibits lagged downgrades and also shows that firms with the same fundamental can have different ratings, both of which are observed in data. We calibrate the model to match a rating accuracy measure and the two distinct average credit spreads at which downgrades and upgrades respectively occur. Policies aiming at improving rating accuracy can either benefit or hurt bondholders. Our calibrated model, however, suggests that such a policy is likely to increase the average value of BBB-rated bonds. The opposite outcome particularly tends to occur to bondholders who hold long-term bonds or face more stringent regulatory pressures.