Information sharing in common agency: when is transparency good?

When should principals dealing with a common agent share their individual performance measures about the agent’s unobservable effort for producing a public good? In a model with two principals who offer linear incentive schemes, we show that information sharing always increases total expected welfare if the principal who is less informed about the agent’s effort also cares more about the agent’s output. If the less informed principal cares somewhat (but not too much) less than the other principal about the agent’s output, information sharing reduces total expected welfare. In our model the efficient information regime emerges as an equilibrium outcome.