A Two-Dimensional Comparison of Signals in an Agency Model

In a general agency model with a risk-averse principal, we compare signals on two dimensions–their efficiencies at providing incentives for the agent's effort and efficiencies at conveying inference of the outcome. In order to tackle the moral hazard problem, the standard first order approach (FOA) is justified by two new sets of sufficient conditions, both of which share the assumption that either the principal is less absolutely risk averse than the agent, or the principal has constant absolute risk aversion. Then, based on the FOA, a strong mean preserving spread-sufficient statistic (MPS-SS) criterion is found to be sufficient to rank arbitrary signals, and an extended informativeness criterion is proven to be both necessary and sufficient for the comparison of inclusive signals