Cheap Talk

Cheap talk is costless, non-binding communication in strategic settings. In game theory, messages are cheap talk when they do not directly affect payoffs — they cost nothing to send or to fake. Crawford and Sobel (1982) proved that even without costs, cheap talk can convey information when sender and receiver interests are partially aligned, though the informativeness decreases as interests diverge.

Cheap talk contrasts with honest signaling, where credibility is guaranteed by differential cost rather than common interest. The evolutionary instability of costless signals in animal communication suggests that cheap talk equilibria may be fragile outside institutional enforcement. In human settings, cheap talk appears in political promises, earnings guidance, and preference surveys — contexts where words lack material consequences but may still coordinate expectations.

Whether cheap talk genuinely conveys private information or merely selects among multiple coordination games remains contested. The distinction matters for behavioral economics, where stated preferences often diverge from revealed ones.

The Structure of Cheap Talk Games

In a cheap talk game, a privately informed sender observes the state of the world and sends a costless message to a receiver, who then takes an action. Because the message is costless and non-binding, the sender has no incentive to tell the truth per se — only to say whatever induces the receiver to take the action the sender prefers. The receiver knows this and updates beliefs accordingly. The equilibrium is a partition of the state space: the sender reveals only which partition element contains the true state, and the receiver takes an action based on this partial information.

The key result of Crawford and Sobel (1982) is that the number of equilibrium partition elements increases with the alignment of preferences. When sender and receiver want the same thing, full revelation is possible. As preferences diverge, the partition coarsens: the sender becomes less informative to prevent the receiver from taking actions that are too far from the sender's ideal. Beyond a critical threshold of divergence, no information is transmitted at all — the "babbling equilibrium," in which the sender's message is pure noise and the receiver ignores it.

Applications and Paradoxes

Cheap talk appears wherever words lack material consequences but may still coordinate expectations. Central bankers issue forward guidance about future interest rates — but the guidance is not a contract, and markets know the bank may renege if conditions change. CEOs provide earnings guidance to investors — but the guidance is not legally binding, and investors know the CEO may sandbag or inflate. Politicians make campaign promises — but promises are not enforceable, and voters know that circumstances (or incentives) may change.

The paradox of cheap talk is that it often works even when it shouldn't. Laboratory experiments show that subjects transmit and respond to cheap talk messages even when the theory predicts babbling. The explanations range from social preferences (people like being honest) to cognitive shortcuts (people take messages at face value because reasoning about strategic manipulation is costly) to institutional habituation (people have learned that cheap talk is often informative in real life, even when the game-theoretic model says it shouldn't be).

Cheap Talk vs. Costly Signaling

Cheap talk and costly signaling represent two poles of the communication spectrum. In cheap talk, credibility comes from common interest: the sender tells the truth because the receiver's optimal response to the truth is close to what the sender wants. In costly signaling, credibility comes from differential cost: the sender tells the truth (or signals type) because lying would be too expensive. Cheap talk is the model of conversation among friends; costly signaling is the model of courtship, credentials, and conspicuous display.

The choice between the two frameworks is not merely a modeling decision; it is a claim about the nature of the social relationship being studied. When economists model firm-manager communication as cheap talk, they assume that the manager's interests are sufficiently aligned with the firm's that costless messages can coordinate behavior. When biologists model animal communication as costly signaling, they assume that interests are in conflict and that only metabolic expense can guarantee honesty. Both assumptions are sometimes right and sometimes wrong, and part of the art of applied theory is knowing which framework fits which context.

Cheap talk is the most underrated force in social life. We coordinate our actions, align our expectations, and build our institutions primarily through costless communication — conversations, emails, norms, narratives — that game theory says should be meaningless. The fact that it is not meaningless suggests that the Crawford-Sobel model, brilliant as it is, captures only a slice of the communication problem. The rest — trust, social pressure, identity, narrative, ritual — operates outside the model's scope but well within the scope of human experience. To understand cheap talk, we may need to stop being game theorists and start being anthropologists.

See also: Bayesian Persuasion, Strategic Information Transmission, Signaling Game, Costly Signaling, Coordination Games, Behavioral Economics