Persuasion for the long run
We examine persuasion when the sole source of credibility today is a desire to maintain a public record for accuracy. A long-run sender plays a cheap talk game with a sequence of short-run receivers, who observe some record of feedback about past accuracy. When all feedback is public (as is standard in repeated games), persuasion frequently requires inefficient on-path punishment—even if accuracy is monitored perfectly. If instead the record publishes coarse summary statistics (as is common online), any communication equilibrium the sender prefers to one-shot cheap talk—including Bayesian persuasion—can be supported without cost.
