Replication Data for "The Limited Value of a Second Opinion"

We experimentally investigate the effect of a second opinion on information transmission. Our design varies the number of senders as well as the alignment and magnitude of senders' biases in a sequential, cheap-talk, sender-receiver game. We find that decision makers do no better when a second opinion is available, irrespective of the alignment or competition between advisers, than when they receive a single opinion. Despite the fact that messaging behavior differs across experimental conditions, receivers successfully extract the same amount of information---an amount greater than what is theoretically predicted. These findings are consistent with senders using a simple strategy of naive exaggeration, with receivers correctly recognizing this and adjusting their behavior accordingly.