Commitment and Randomization in Communication

When does a Sender, in a Sender-Receiver game, strictly value commitment? In a setting with finite actions and finite states, we establish that, generically, Sender values commitment if and only if he values randomization. In other words, commitment has no value if and only if a partitional experiment is optimal under commitment. Moreover, if Sender's preferred cheap-talk equilibrium necessarily involves randomization, then Sender values commitment. We also ask: how often (i.e., for what share of preference profiles) does commitment have no value? For any prior, any independent, atomless distribution of preferences, and any state space: if there are $\left|A\right|$ actions, the likelihood that commitment has no value is at least $\frac{1}{\left|A\right|^{\left|A\right|}}$. As the number of states grows large, this likelihood converges precisely to $\frac{1}{\left|A\right|^{\left|A\right|}}$.