Arjada Bardhi, Nina Bobkova
Anecdotes pervade public discourse. In a sender-receiver framework, we formalize why anecdotes persuade even when an aggregate statistic—rather than individual cases—is what matters. A single anecdote reveals information about similar cases, shifting beliefs about the statistic through correlation. The optimal anecdote balances reach—how correlated it is to other cases—against rarity—how surprising good news would be. Optimal anecdote selection follows a two-stage procedure: First identify cases persuasive enough upon good news, then choose the case most likely to deliver such news. The sender's optimal, receiver's preferred, and most informative anecdotes can diverge markedly.