Decision errors and threshold public goods: Using logit equilibrium to model coordination failure

The rise of online threshold fundraising offers a novel opportunity for the private provision of public and information goods. However, the underlying coordination problem may lead to inefficient allocations. Online mechanisms have two unique features that are not well-understood: (i) uncommonly large groups and (ii) little information regarding others’ incentives. Intuitively, increasing group-size or payoff uncertainty might exacerbate coordination failure; although, it is not obvious how to predict such effects in practice. I conduct a two-by-two experiment varying group-size and payoff uncertainty, and I use the data to estimate a logit quantal response equilibrium model. This approach captures a great deal of observed behavior at the individual and group level, and it accounts for the sign of all significant treatment effects. The model underpredicts the average level of contributions and the success rate. One way to correct for this bias is through the addition of social preferences, which is consistent with the findings of many public goods experiments.