Financial covenants, which use financial information to determine control rights in corporate loan contracts, are frequently amended (Roberts, 2015). Despite the strong evidence, prior literature has not found an explanation for the high frequency. In this paper, I offer a model of debt covenant amendments and structurally estimate it using novel data on amendments of financial covenants in U.S. companies' loan contracts. The model offers three explanations for why financial covenants get amended: (1) the degree of contractual incompleteness is high, (2) non-contractible information is highly predictive of the future, and (3) the costs of amendments are small, while the costs of misallocating control rights are high. The estimation results mostly support the first and the third explanations. Financial covenants are amended because contracting parties indeed face a high degree of contractual incompleteness, because the costs of amendments are indeed low compared to the costs of misallocating control rights, but not necessarily because non-contractible information is highly predictive of the future.