We study the wage negotiation model of Haller and Holden (1990) and Fernandez and Glazer (1991) under the "Good Faith Bargaining" (GFB) rule, where a party may not demand more than it has previously demanded. The GFB rule significantly restricts feasible strategies, but at the same time, makes the game non-stationary and the analysis complicated. We introduce a state-dependent backward induction that generalizes Shaked and Sutton (1984) to characterize the equilibrium payoffs. We find that the GFB rule eliminates the union's credibility to strike. Without the strikes, the union's strategic opportunities during disagreement disappear, so that there is a unique equilibrium. This uniqueness contrasts sharply with the multiple equilibrium outcomes that obtain when no GFB rule is imposed.