Long Cheap Talk 1

STRATEGIC INFORMATION TRANSMISSION has been studied in economic theory for over a quarter of a century. Most formal models in this area allow for at most one message from each player. Yet in practice, as in negotiation or bargaining, protracted exchanges of messages are often observed. Does this make economic sense? Can a long exchange convey substantive information that cannot be conveyed by a single message? Here we examine this question in the context of “cheap talk.” The answer is “yes”: Long cheap talk may lead to outcomes preferred by all players to those achievable with single messages. We will characterize all the equilibrium outcomes to which it can lead, in any two-person game in which one player is initially better informed than the other. Cheap talk is just that: cheap—neither costly nor binding; and talk— not some roundabout form of communication, like mediation. Unlike “signalling,”2 cheap talk—plain conversation—is “payoff-irrelevant;” there is no “credibility cost.” The players don’t strike, don’t get educated, and don’t issue guarantees; they simply talk. They may or may not tell the truth, and may or may not believe each other. To be sure, that there is no credibility cost does not mean that there is no credibility; depending on the circumstances, the cheap talk may itself create positive motivation for the players to believe each other (Examples 2.1 and 2.2 below). It is this motivation that forms the crux of our analysis. In a sense, cheap talk is communication in its purest and simplest form: purest in that there is no direct impact on the payoff, and simplest in that there is no intermediary. The literature on cheap talk addresses two issues, almost diametrically opposite. One is how cheap talk restricts the set of outcomes—equilibrium refinement; see Section 10. The other is how cheap talk expands the set of outcomes.