I study a repeated game where a principal hires a risk-averse agent through long-term contracts on output. The agent’s effort is observable but non-contractible and is a basis for a self-enforcing agreement (relational contract) between the players: after observing effort and output the principal can pay a voluntary bonus and attempt to renegotiate the contract through a take it or leave it offer. In equilibrium renegotiation can create a Pareto improvement even though every outcome is observable. The benefit comes from designing continuation contracts which allow the players to initiate harsh punishments if one of them deviates from the relational contract. After the agreed upon effort and bonus are observed, these contracts are renegotiated to new contracts which are better suited to supporting efficiency on the equilibrium path. In a model without limited liability for the principal, the first best outcome can always be achieved by constructing continuation contracts with arbitrarily severe punishments for the principal which makes high bonuses credible, creating incentives for the agent without distorting his insurance. Two-period contracts are sufficient to implement the first best and the result is independent of the patience of the players and the output technology. Renegotiation is essential, as long as the players are not too patient. In a model with limited liability the harshest punishments amounting to the outside options can be invariably achieved through continuation contracts which are renegotiated on path. This facilitates an equivalence to a one-period contracting problem.
Kostadinov, Rumen, Renegotiation of Long-Term Contracts with Implicit Incentives (September 16, 2017).