This paper proposes a new approach to equilibrium selection in repeated games with transfers, supposing that in each period the players bargain over how to play. Although the bargaining phase is cheap talk (following a generalized alternating-offer protocol), sharp predictions arise from three axioms. Two axioms allow the players to meaningfully discuss whether to deviate from their plan; the third embodies a “theory of disagreement”—that play under disagreement should not vary with the manner in which bargaining broke down. Equilibria that satisfy these axioms exist for all discount factors and are simple to construct; all equilibria generate the same welfare. Optimal play under agreement generally requires suboptimal play under disagreement. Whether patient players attain efficiency depends on both the stage game and the bargaining protocol. The theory extends naturally to games with imperfect public monitoring and heterogeneous discount factors, and yields new insights into classic relational contracting questions.

We study relational contracting and renegotiation in environments with external enforcement of long-term contractual arrangements. A long-term contract governs the stage games that the contracting parties will play in the future (depending on verifiable stage-game outcomes) until they renegotiate. In a contractual equilibrium, the parties choose their individual actions rationally, jointly optimize when selecting a contract, and exercise their relative bargaining power. Our main result is that in a wide variety of settings, the optimal contract is semi-stationary, with stationary terms for all future periods but special terms for the current period. In each period the parties renegotiate to this same contract. For example, in a simple principal-agent model with a choice of costly monitoring technology, the optimal contract specifies mild monitoring for the current period but intense monitoring for future periods. Because the parties renegotiate in each new period, intense monitoring arises only off the equilibrium path after a failed renegotiation. (JEL C73, C78, D23, D86)