I study the impact of ''say on pay'' (SoP) on the compensation decisions and the structure of the board of directors (BoD) in a setting where the CEO has the real authority over the composition of the BoD. The CEO's authority arises endogenously from an informational advantage about individual board members' contribution to firm value and allows her to establish a dependent BoD. Shareholders approve the CEO's director slate because they can only control the level of board dependence but not the board's contribution to firm value. In this setting, SoP has two effects. On the one hand, it prompts a BoD with a given dependence level to reduce the CEO's bonus. On the other hand, it allows the CEO to extract the rent generated by the improved compensation policy and to establish a more dependent BoD. In equilibrium the board becomes more dependent from the CEO and pays her a higher bonus for the same performance. This outcome can only be avoided if the CEO is restricted in her ability to adjust the board composition. Motivated by existing differences in SoP design, I also analyze the consequences of a binding and a pre-contractual vote. I find that a binding vote creates a moral hazard problem on the part of the firm's shareholders if the vote takes place after the agent has supplied her effort. Its consequences critically depend on the legal protection standard of the CEO. Whenever the shareholders can enforce a retroactive bonus cut, the allowable amount of the bonus reduction determines whether or not SoP improves the efficiency of the pay process or diminishes firm value. I show that the moral hazard problem can be avoided by a pre-contractual vote. If the vote is binding, SoP can improve the efficiency of the compensation arrangement and effectively reduce the equilibrium level of board dependence without impairing the CEO's effort incentives.