Speaker: Giovanna Devetag (University of Perugia), joint work with: Andreas Ortmann
For information: Mikhail Anufriev.
Coordination games with Pareto-ranked equilibria have attracted major theoretical attention over the past two decades, as path-breaking experimental studies (Van Huyck, Battalio and Beil, 1990, 1991; Cooper, DeJong, Forsythe and Ross, 1990, 1992) were widely interpreted as suggesting that coordination failure (i.e., coordination on Pareto-inferior equilibria) is a common phenomenon in the laboratory. In Devetag and Ortmann (2006), after a careful review of the literature, we have argued that coordination failures are much less common than is widely perceived (e.g., Ochs 1995 and Camerer 2003). In fact, we have argued that it is by now well understood how coordination successes can be engineered. We also formulated specific conjectures about the impact of deviation costs and loss aversion.
In the present manuscript, we focus on the median action game first studied by Van Huyck, Battalio and Beil, 1991, to experimentally test these conjectures. Early results on the median action game had shown both a high frequency of coordination failure and strong history-dependence, in that the last round median was always equal to the first round median in all treatments. In our experiments we test the robustness of these results by manipulating two variables related to the payoff function: the presence vs. absence of negative payoff cells in the earnings table, and the presence of linear vs. nonlinear deviation costs, i.e., costs of overshooting relative to some equilibrium configuration.
Our results show that, by and far, our conjectures are verified: in particular, when negative payoffs are absent and deviation costs are linear, we observe a higher incidence of coordination success.