The public bureaucracy is a puzzle. How is it that an organizational form that is so widely used is also believed to be inefficient - both in relation to a hypothetical ideal and in comparison with private bureaucracies? This article examines public bureaucracy through the lens of transaction cost economics, according to which the public bureaucracy, like other alternative modes of governance, is well suited to some transactions and poorly suited to others. Rather than proceed in a completely general way, I focus on what James Q. Wilson describes as 'sovereign transactions', of which foreign affairs is an example. I ask what it is that distinguishes sovereign transactions, after which I compare the efficacy of public and private bureaucracies for managing such transactions. I conclude that there is an efficiency place for public bureaucracy, but that all modes of governance (markets, hybrids, firms, regulation), of which public bureaucracy is one, need to be kept in their place. I further observe that public bureaucracies are not all of a kind and that differences between them need to be distinguished.