Relational incentive contracts with productivity shocks

This paper extends Levin's (2003) relational contract model by having not only the agent's cost of effort (agent's type), but also the value of that effort to the principal (principal's type) subject to i.i.d. shocks. When optimal effort is fully pooled across agent types for multiple principal types, it is also pooled across those principal types. When optimal effort separates some agent types for multiple principal types, efforts of those agent types may be separated across principal types. But then, somewhat perversely, some agent type's effort is decreasing in the principal's value of effort. When agent type is uniformly distributed, that applies to agent types with lower effort cost, so reducing the difference in effort between low and high effort cost types. This result extends to the principal's type being observed only by the principal if the marginal cost of effort to the agent is sufficiently convex.