Cheap Talk is not Cheap: Free versus Costly Communication

How do people communicate their intentions before making decisions that affect others? This paper explores this question in a setting where two players must independently choose between two actions, knowing that their choices depend on each other but without being able to directly coordinate. Players fall into two types: some always prefer the “safe” low action regardless of what the other does, while others do best by matching their partner’s choice — and both choosing the high action is their ideal outcome. Before deciding, players can send signals about what they intend to do. We compare three settings: no communication, costless messages (cheap talk), and costly messages where sending a signal requires paying a fee. Standard theory predicts that costly signals should work better, since only genuinely cooperative players would pay to signal. But in our experiment, the opposite happened: cheap talk led to better outcomes for cooperative players, even beating the best outcome theory predicts. Why? Because roughly half of the “safe” players refused to send misleading signals — they were unwilling to make promises they wouldn’t keep. This reluctance made cheap talk surprisingly informative. With enough “promise-keepers” in the population, the higher rate of signaling under cheap talk more than compensates for its lower credibility, allowing cooperative players to coordinate on the mutually beneficial outcome more often. The findings suggest that social norms around honesty can make costless communication more effective than costly signaling mechanisms — a result with implications for the design of communication institutions.

CEPR DP 15843