The persistance of "bad" precedents and the need for communication : a coordination experiment

Precedents can facilitate successful coordination within groups by reducing strategic uncertainty, but they may lead to coordination failure when two groups with diverging precedents have to interact. This paper describes an experiment to explore how such coordination failure can be mitigated and whether subjects are aware of it. In an initial phase, groups were able to establish a precedent in a repeated weakest-link game, and in a second phase two groups with different precedents are merged into a larger group. As expected, this leads to coordination failures. Unlike most of the previous literature, subjects could endogenously choose to communicate in the merged group for a small fee. The results suggest that communication can mitigate the coordination failure in the merged group and, in most cases, leads to efficient coordination. However, subjects in particular from groups with an efficient precedent in the initial phase are inattentive to the potential coordination failure and choose not to communicate. This can have profound consequences since groups who fail to implement communication are unable to achieve effcient coordination in the second phase. The results may be useful for the understanding of how groups learn to solve coordination problems from past coordination success or failure. -- coordination ; precedent ; costly communication ; cheap talk