Modelling contributions in public good games with punishment

Theoretical models have had difficulties to account, at the same time, for the most important stylized facts observed in experiments of the Voluntary Contribution Mechanism. A recent approach tackling that gap is Arifovic and Ledyard (2012), which implements social preferences in tandem with an evolutionary learning algorithm. However, the stylized facts have evolved. The model was not built to explain some of the most important findings in the public good games recent literature: that altruistic punishment can sustain cooperation. This paper extends their model in order to explain such recent findings. It focuses on fear of punishment, not punishment itself, as the key mechanism to sustain contributions to the public good. Results show that our model can replicate both qualitatively and quantitatively the main facts. Data generated by our model differs, on average, in less than 5% compared to relevant experiments with punishment in the lab.