Abstract:
The paper addresses a classic unsolved question of information economics. How can a principal best motivate a limited-liability agent whose action she observes with N(0,σ) measurement error? We characterise the solution for a wide class of cases. The solutions are easily computable. Some interesting comparative statics arise: for instance agents often work harder when there is moral hazard, than they would in the first-best. The methods generalise to families of distributions other than N(0,σ).