We investigate whether "binding agreements" offer a solution for solving the social dilemma that arises in the presence of pure public goods. A binding agreement is defined as an agreement of agents who agree to contribute a fixed level of their endowment to the provision of the public good. In our setting, the individual level of contribution to the public good in theory increases in the size of the agreement and the complete agreement is always the socially optimal agreement structure. Agreements form sequentially and the equilibrium outcome is an asymmetric structure, which consists of two agreements of which the smaller one forms first. Our experimental results show that the formation of binding agreements frequently leads to inefficient agreement structures, with even a lower performance than at the theoretical equilibrium. We explore two possible explanations for such an unexpected outcome: inequality avoidance and myopic best reply.