| Abstract: | This paper describes an experiment in which subjects, acting as division managers, exchanged privately held information before making intrafirm investment decisions. Social efficiency required that managers honestly disclose their private information, but managers had individual incentives to send biased messages. These features of the model created an important role for ex post verification, the main manipulation in the experiment. The matching protocol was also manipulated, using both random and continuous matching of subjects. This second manipulation was intended to examine whether an important institutional attribute — the frequency of interaction — would affect the usefulness of verification. The results of the experiment indicate that verification significantly increased the relative frequency of honest messages and the level of social efficiency. However, the improvements from verification were greater in settings where subjects did not interact repeatedly. The data also indicate that, in the continuous matching treatments, responses depended on the history of behavior of the message sender. However, this behavior was not observed in the random matching treatments. Thus, both the efficacy of verification and the extent of reputation formation depended on the institutional setting. |
