These data were gathered to study coordination games in which individuals have conflicting preferences. We propose the actors’ ability to vary behavior when interacting with different partners (partner-specific behavior) as a mechanism facilitating coordination in situations where actors have different preferences. We assume that actors are connected through a network and play coordination games with their neighbors in the network. Actors may or may not behave partner-specifically and actors’ preferences for outcomes in the game may differ. Both actors prefer coordination over miscoordination.
Actors are arranged on a network. If actors i and j are connected in the network, they play the coordination game. This network remains unchanged over the repeated game. Symmetric networks are used: all actors within one network have the same number of neighbors. The actors have global information regarding their own and the preferences of all other actors in the network and regarding the ability of actors to behave partner-specifically or not. Actors receive also information about the behavior of all actors in the network.
Subjects in our experiment played repeated coordination games for 20 periods among a group of four actors. They received points dependent on the decisions they made. At the end of the experiment, points were exchanged for monetary earnings. We distinguish three versions of the game. (1) Actors cannot behave partner-specifically, i.e. each actor has to choose the same behavior with all neighbors. (2) Actors can behave partner-specifically, i.e. each actor can differentiate behavior in interactions with different neighbors. (3) Actors can buy the ability to behave partner-specifically. We used different networks to create variance with respect to the heterogeneity in preferences within networks as well as the number of heterogeneous dyads in networks, while keeping the structural positions within networks constant.
The experiment was programmed and conducted with z-Tree software. Subjects were recruited amongst students
at Utrecht University. The total number of subjects is 144. There were 68 (47.2%) male subjects. The age of the subjects ranged from 18 to 56 with an average age of 24.
Results show that whether partner-specific behavior is disadvantageous or advantageous for coordination depends on the distribution of preferences in the network. Moreover, subjects seem unable to foresee when partner-specific behavior is disadvantageous, since they invest in partner-specific behavior also when this does not pay off.
