Credibility in Persuasion: A Laboratory Experiment

We experimentally examine whether partial commitment to the message distribution can substitute for full commitment to the information structure in Bayesian persuasion. Theory predicts that when the players’ marginal incentives are aligned, an equilibrium with the same level of information transmission as under full commitment can be sustained even under partial commitment, whereas when marginal incentives are misaligned, no information can be credibly transmitted. Our central finding is that, even when marginal incentives are aligned, information transmission is substantially lower under partial commitment than under full commitment: partial commitment does not fully substitute for full commitment in practice. When marginal incentives are misaligned, information transmission declines further under partial commitment, as theory predicts. We also find that misaligned marginal incentives reduce information transmission even under full commitment, where the standard theory predicts no such effect. Our analysis of sender behavior shows that, in all of these cases, senders transmit less information than the theoretical benchmark implies, suggesting that sender undercommunication plays an important role in the lower information transmission we observe.