Conditional cooperation in network public goods experiments

This study investigates the pattern of contribution decisions in a network public goods game. In this game, each player’s payoff depends only on his own contribution and the contributions of his immediate neighbors in a circle network. As in the standard public goods game, we find substantial heterogeneity in behavior across subjects, including both unconditional free-riding and full cooperation, as well as conditional cooperation. We first examine the impact of different information conditions on conditional cooperation. At the aggregate level, we find that players who observe average payoff information about others contribute significantly less than those who observe average contribution information. We then investigate the extent to which conditional cooperators facilitate the spread of cooperation and free-riding behavior across the network. In groups with a single free-rider type, we show that individual contributions decay faster for players who are closer in the network to the free rider. On the other hand, in groups with a single unconditional full contributor type, players do not respond by converging to full cooperation. Instead, we find that proximity to the unconditional full contributor seems only to mitigate (or delay) the typical decline in contributions over time. These contrasting effects are consistent with the widespread claim that conditional cooperation is imperfect, or exhibits a self-serving bias.