Network effects on strategic interactions: A laboratory approach

- One's choices in herding and shirking games are inversely affected by her degree.
- The observed local [global] network effect is stronger in shirking [herding] games.
- The laboratory observations are captured by a hybrid learning model in the network.
- Individual behavior is mainly driven by the adaptation to realized neighbor action.

We examine the effect of interaction structure (network) on two classes of collective activities, herding and shirking, respectively referring to the situation where a player's incentive to take a certain action increases and decreases if more of her network neighbors follow the same action. In our experiment, we find that subjects do not act according to theoretical equilibrium, and their frequencies of making the socially beneficial choice in herding and shirking games are inversely influenced by the number of network neighbors they have. Moreover, the observed local network effect is stronger in shirking games, while the global network effect is more significantly present in herding games. We explain the behavioral regularities through a hybrid learning model, which extends SEWA learning into a network context. As such, our learning model provides a foundation for the observed dynamics, disequilibrium behavior, as well as the local and global network effects.