See no evil: Information chains and reciprocity

We study experimentally voluntary contributions to public goods when none, some, or all previous decisions are observable. When agents observe previous moves, they tend to condition their cooperation on observed cooperation. This leads to two effects of increased transparency: on the one hand, early movers are more likely to cooperate in order to encourage those who observe them to cooperate. On the other hand, as transparency increases, later movers are less likely to cooperate because they are more likely to observe defections and defect in response. With increasing returns to scale, where the effect of one agent's contribution is larger as more agents contribute, an information chain is as effective in inducing cooperation as full transparency. In a linear public good, where agents lose in monetary terms by contributing to the public good, information chains induce higher cooperation in early movers compared to a no-transparency treatment and in late movers compared to a full-transparency treatment. Thus, partial information can be used to balance the positive and negative effects of transparency.