Predicting efficiency in threshold public good games: a learning direction theory approach

In this paper we propose a tractable model of behavior in threshold public good games. The model is based on learning direction theory. We find that individual behavior is consistent with the predictions of the model. Moreover, the model is able to accurately predict the success rate of groups in providing the public good. We apply this to give novel insight on the assurance problem by showing that the problem (of coordinating on the inefficient equilibrium of no contributions) is only likely with a relatively low endowment. In developing the model we compare and contrast best reply learning and impulse balance theory. Our results suggest that best reply learning provides a marginally better fit with the data.