Discrete public goods: Contribution levels and learning as outcomes of an evolutionary game

This paper examines the learning dynamics of boundedly rational agents, who are asked to voluntarily contribute to a discrete public good. In an incomplete information setting, we discuss contribution games and subscription games, the latter including a money-back guarantee in case of provision failure. The theoretical results on myopic best response dynamics implying striking differences between strategies played in the two games are confirmed by simulations, where the learning process is modeled by an Evolutionary Algorithm. We show that the contribution game even aggravates the selective pressure leading towards the non-contributing equilibrium, thereby supporting results from laboratory experiments. In contrast to this, the subscription game removes the 'fear incentive', implying a higher percentage of successful provisions over time.