|2002/09||LEM Working Paper Series|
Coordination and Self-Organization in Minority Games: Experimental Evidence
| Giulio Bottazzi and Giovanna Devetag |
Minority Game, Experimental Economics, Emergent Coordination,Aggregate Efficiency.
This work presents experimental results on a coordination game inwhich agents must repeatedly choose between two sides, and apositive fixed payoff is assigned only to agents who pick theminoritarian side. The game presents a variety of asymmetric purestrategy equilibria, and a unique symmetric mixed-strategyequilibrium in which agents randomize between the two sides atevery stage. The game reflects some essential features of thoseeconomic situations in which positive rewards are assigned toindividuals who behave in opposition to the modal behavior in apopulation. We conduct laboratory experiments in which stationarygroups of five players play the game for 100 periods, andmanipulate two treatment variables: the amount of "memory" Mthat players have regarding the game history (i.e., the length ofthe string of past outcomes that players can see on the screenwhile choosing), and the amount of information about otherplayers' past choices: in the aggregate information treatment,players only know which side was the minority side at each period,while in the full information treatment players have informationregarding the entire distribution of choices in the group at eachround. We first analyze aggregate results in terms of both"allocative" and "informational" efficiency. We thenanalyze individual behavior in the game as compared to thetheoretical benchmark provided by the mixed strategy equilibriumsolution. Our results show that, first, both allocative andinformational efficiency are higher on average than the benchmarkvalue corresponding to the mixed strategy equilibrium in alltreatments, suggesting that a quite remarkable degree ofcoordination is achieved; second, providing players with fullinformation about other players' choice distribution does notappear to improve efficiency significantly. At the individuallevel, a substantial portion of subjects exhibit `inertial'behavior, i.e., the tendency to replicate their previous roundchoice with a higher frequency than the one prescribed byrandomizing behavior, and such inertia seems to be enhanced ratherthan decreased by a full information treatment.