Comparisons of Experiments in Moral Hazard Problems

Abstract

I use a novel approach to compare information in several classes of moral hazard problems: implementability, cost under risk neutrality and limited liability, and cost facing an agent with a general preference for money. Incentives in moral hazard problems are determined by the agent's state-dependent utility. Motivated by this observation, I define three nested geometric orders on information: the column space, the conic span, and the zonotope orders. Each order admits four equivalent characterizations: (i) inclusion of feasible state-dependent utility sets, (ii) dominance in the corresponding class of moral hazard problems, (iii) matrix factorizations, (iv) posterior belief distributions. In particular, the orders apply to both the classic and the flexible moral hazard problems, providing a unified framework to compare information.