Abstract.  Peer-to-peer networks introduce a new area of interplay between computer science and economics. Designers of such systems must firmly understand the incentives, preferences, and decision space of participating agents in order to decide the policies and make the system function as well as possible. This paper models the economic behavior of agents in a peer-to-peer storage network. From the model, it becomes clear that agents have single-peaked preferences for a system-wide parameter that defines the reliability of the storage network. Consequently, the system designer may implement a mechanism to elicit opinions for this parameter (knowing that they will be truthfully revealed) and set the system-wide value to some socially optimal level, or agents with similar preferences may cluster together to form a p2p network closer to their preferences.