Do Actions Speak Louder Than Words? An Experimental Comparison of Observation and Cheap Talk

How do individuals achieve \good outcomes" in one{shot strategic situations? One possibility is that they engage in some kind of preplay communication|cheap talk|in which they endeavor to convince one another of the actions they intend to play. Another, less explored, possibility is that individuals take account of their knowledge of the past behavior of others when deciding which actions to play. While these two possibilities have been considered separately, there has been no research examining the relative e cacy of cheap talk and observation for the achievement of good outcomes. This paper reports the results from an experiment with human subjects that allows for such a comparison. The e ects of cheap talk and observation of past actions are compared with each other, and with a control, in which neither cheap talk nor observation is allowed. We consider three di erent 2 2 games and explain why cheap talk or observation is likely to be the more e ective device for achieving good outcomes in each game. The experimental evidence suggests that both cheap talk and observation make cooperation and successful coordination more likely and increase payo s relative to the control. The relative success of cheap talk versus observation in achieving such good outcomes depends on the game played, in accordance with our predictions. We also nd that the signals players send are informative in the sense that they are correlated with their eventual actions, and that receivers of signals take this fact into account by conditioning their actions on the signal they receive. The results of this experiment can be used to extend game{theoretic models of how individuals make use of the di erent types of information available in strategic environments. Funding for this project was provided by National Science Foundation grants SBR{9809622 and SBR{9813622. The raw data from the experiments we report in this paper are available from the authors upon request. We thank Nicole Bouchez, Catherine Eckel, Ann Gillette, Jack Ochs, Rick Wilson, participants at several conferences and seminars, two anonymous referees, and an associate editor for helpful comments and suggestions on earlier drafts. Any remaining errors will be useful in predicting our future actions.