We consider a model in which a security issuer can manipulate information observed by a credit rating agency (CRA). We show that stricter screening by the CRA can sometimes lead to increased manipulation by the issuer. Accounting for the issuer’s behavior pulls optimal CRA screening toward the extremes of laxness or stringency. Surprisingly, an improvement in prior asset quality can result in more rating errors. In a two-period version of the model, stricter screening can result in more short-run rating errors. Our results suggest complex interplay between issuer and CRA behavior, complicating the evaluation of CRA policy effectiveness.