We present a new and simple mechanism for repeated public good environments. In the Binary Conditional Contribution Mechanism (BCCM), every agent's message has the form, “I am willing to contribute to the public good if at least k agents contribute in total.” This mechanism offers agents risk-free strategies, which we call unexploitable. We prove that if agents choose unexploitable messages in a Better Response Dynamics model, all stable outcomes of the BCCM are Pareto efficient. We conduct a laboratory experiment to investigate whether observed behavior is consistent with this prediction. Subjects play the BCCM in an environment with complete information and homogeneous valuations or in a second environment with incomplete information and heterogeneous valuations. In both cases all stable outcomes in the experiment are in line with the prediction of the dynamic model. Furthermore, comparison treatments with the Voluntary Contribution Mechanism show that the BCCM leads to significantly higher contribution rates.