We introduce a model of a rational credit card user's rather complex usage choices and develop an empirical framework to test its predictions. Employing a large national database of U.S. card accounts, we estimate how prices impact card usage and find that price effects are mostly well explained within our model. An exception is less borrowing in response to declining late-fees among low credit-score (subprime) users. Extension of our model based on "focusing theory" predicts this behavior. It also implies substantial indirect benefits of the CARD Act's late-fee cap due to subprime users re-focusing toward reducing their debt.