The assumption of full commitment to an incentive scheme was already discussed in section 2.12 in the case of adverse selection. This issue is also quite important under moral hazard. For instance, to induce a positive effort level the principal must let the risk-averse agent bear some risk. However, once this effort is sunk, and before uncertainty is resolved, the principal would like to offer more insurance to the agent to avoid paying an excessive agency cost. For this reinsurance stage to have any impact, the principal must be aware, maybe through direct observation of the effort itself or by indirectly getting a signal correlated with the effort, that effort has already been performed. Of course, the renegotiation stage would be perfectly anticipated by the rational agent at the time of exerting effort. Renegotiation is unlikely to lead to complete insurance ex post, because the agent would then have no incentive to exert effort in the first place.
Fudenberg and Tirole (1990) showed how the possibility of renego- tiation induced a mixed strategy in the effort provision of the agent. The complexity of the model comes from the fact that the agent’s actual choice of effort becomes an endogenous adverse selection variable at the renegotiation stage. This creates an inefficiency. This inefficiency can be avoided if the informed party makes the renegotiation offer, as in Matthews (1995), who allowed renegotiation on the equilibrium path, or Ma (1994), who focused on renegotiation-proof contracts. Finally, Hermalin and Katz (1991) showed that the first-best outcome could be implemented when the principal observes the nonverifiable effort of the agent before renegotiation, because then renegotiation takes place under complete information.
Source: Laffont Jean-Jacques, Martimort David (2002), The Theory of Incentives: The Principal-Agent Model, Princeton University Press.