Communication in Private-Information Models: Theory, Computation, and an Application

Communication and no-communication versions of a two-stage principal-agent model are compared. The models contain a risk-averse agent and two sources of private information, a shock to preferences followed by a productive action. Both models are formulated as linear programs, which are then used to compute solutions to examples. In simple cases, a sufficient condition for communication to be valuable is derived. In these cases, communication improves risk-sharing in bad states of the world. In more complicated cases, computed examples demonstrate how communication may also alter labor supply. Further examples demonstrate how action and consumption lotteries may separate agents by their shock. Survey evidence and Þeld work on contractual arrangements between farmers and landowners in an Indian village is used to motivate the class of models. Suggestive evidence on communication from the same sources is also described.