This paper studies the costs and benefits of delegating decisions to superiorly informed agents relative to the use of rigid, non discretionary contracts. Delegation grants some flexibility in the choice of the action by the agent, but also requires the use of an appropriate incentive contract so as to realign his interests with those of the principal. The parties' understanding of the possible circumstances in which actions will have to be chosen and their attitude towards risk and uncertainty play then an important role in determining the costs of delegation. The main focus of the paper lies indeed in the analysis of these costs and the consequences for whether or not delegation is optimal. We determine and characterize the properties of the optimal flexible contract both when the parties have sharp probabilistic beliefs over the possible events in which the agent will have to act and when they only have a set of such beliefs. We show that the higher the agent's degree of risk aversion, the higher the agency costs for delegation and hence the less profitable is a flexible contract versus a rigid one. The agent's imprecision aversion in the case of multiple priors introduces another, additional agency costs; it again implies that the higher the degree of imprecision aversion the less profitable flexible contracts versus rigid ones. Even though, with multiple priors, the contract may be designed in such a way that principal and agent end up using 'different beliefs' and hence engage in speculative trade, this is never optimal, in contrast with the case where the parties have sharp heterogeneous beliefs.
