The field of cultural economics has devoted substantial attention to the positive and normative analysis of the willingness of citizens to fund art through the public sector. But there has been little economic analysis of the efficiency properties of the various possible methods of organizing public funding of the arts. Recent work on transaction cost models of organizations and politics can be fruitfully applied to the study of publicly funded arts agencies. In particular, the brief history of the National Endowment for the Arts (NEA) provides a vivid illustration of the problem of optimal degrees of political monitoring and statutory control of agencies. The special features of the art world, in particular the status and knowledge of art experts and the lack of a clear sense of mission as to the purpose of publicly funded art, and a rapid change over the past four decades in what artists actually do, have led to changes in the statutory control of the NEA and in the way it is monitored. Examples include the changes in the composition of the National Council on the Arts, the end of grants to individual artists (with the exception of writers), and proportionately large changes in the NEA’s budget allocation. This paper uses a transaction cost approach to link the changes in the art world and the changes in the oversight and procedures of the NEA. Framing the issue as a public administration problem, the paper highlights the challenges in using expert advisory panels when the goals of the panels are uncodifiable.