We provide a new framework for using text as data in empirical models. The framework identifies salient information in unstructured text that can control for multidimensional heterogeneity among assets. We demonstrate the efficacy of the framework by reexamining principal-agent problems in residential real estate markets. We show that the agent-owned premiums reported in the extant literature dissipate when the salient textual information is included. The results suggest the previously reported agent-owned premiums suffer from an omitted variable bias, which prior studies incorrectly ascribed to market distortions associated with asymmetric information.