[In the usual agency analysis of moral hazard, the principal is endowed with a production process and must hire an agent to manage it. Output depends on the agent's effort, the production process supplied by the principal, and some random state realization. A common assumption is that the output of the production process is costlessly observable and sufficiently informative about the agent's effort to warrant using it for contracting. However, this assumption may not be descriptive of a large number of settings. In fact, it is hard to conceive of many production settings in which the gathering and reporting of any information is totally free and independent of costly design decisions. For example, if the agent produces some product, it is not costless to monitor the number or the quality of the units produced; someone must be paid or a machine must be purchased to do so. In addition, some opportunity cost may have to be incurred to rearrange the production process to facilitate the assessment of the agent's work. In this article, we consider the implications of relaxing the assumption that output is costlessly observable and contractible, by analyzing the optimal design of a costly unconditional monitoring system. We characterize the monitoring system in terms of the Type I error associated with the obedient action and the Type II errors associated with the disobedient actions. We find that it is always optimal to design the monitoring system such that the Type I error is smaller than the Type II error for any disobedient action. Further, we find that as the costliness of the monitoring system increases, the Type I error increases monotonically, while the Type II error initially decreases and then increases.]

Analyzes the design of an unconditional monitoring system in agencies. Value of monitoring; Characterization of monitoring system in terms of error associated with the obedient and disobedient actions; Optimal design of the costly unconditional monitoring system.