Cheap Talk

In the context of games of incomplete information, the term “cheap talk” refers to direct and costless communication among players. Cheap-talk models should be contrasted with more standard signalling models. In the latter, informed agents communicate private information indirectly via their choices— concerning, say, levels of education attained— and these choices are costly. Indeed, signalling is credible precisely because choices are di¤erentially costly— for instance, high productivity workers may distinguish themselves from low productivity workers by acquiring levels of education that would be too costly for the latter. The central question addressed in cheap-talk models is the following: How much information, if any, can be credibly transmitted when communication is direct and costless? Interest in this question stems from the fact that with cheap talk, there is always a "babbling" equilibrium in which the participants deem all communication to be meaningless— after all it has no direct payo¤ consequences— and as a result, no one has any incentive to communicate anything meaningful. It is then natural to ask if there are also equilibria in which communication is meaningful and informative. We begin by examining the question posed above in the simplest possible setting: there is a single informed party— an expert— who o¤ers information to a single uninformed decision maker. This simple model forms the basis of much work on cheap talk and was introduced in a now classic paper by Crawford and Sobel (1982). In what follows, we …rst outline the main …nding of this paper— while there are informative equilibria, these entail a signi…cant loss of information. We then examine various remedies that have been proposed to solve (or at least alleviate) the “information problem.”