What makes cheap talk effective? Experimental evidence

In some environments, a player only learns the choice of another player if he or she undertakes a risky choice. While costless preplay communication (cheap talk) has been found to be effective in experimental coordination games, participants have typically learned both own payoffs and the other player's action. Are both of these components necessary for cheap talk to be effective? In our 2 x 2 stag hunt game, the safe choice always yields the same payoff, so that information about payoffs does not always identify the other player's action. We vary whether information is provided about the other person's play, and whether costless one-way messages can be sent before action choices are made. We find that information provision about the other person's play increases coordination when there are messages, but otherwise has no effect. (C) 2004 Published by Elsevier B.V.