Institution formation in weakest-link games

We study the role of endogenous formation of institutions in overcoming coordination failures in weakest-link games with fixed neighborhoods. In our setting, institutions are weak and only form and make decisions by unanimity. Experimental results show that such institutions are formed and mitigate the coordination problem, raising equilibrium provision levels, but falling short of providing Pareto-optimal contributions. Given the multiplicity of Nash equilibria in weakest-link games, we consider several equilibrium refinements that allow for (small) errors by individuals. Without institutions, risk dominance and the Quantal Response Equilibrium (QRE) with (almost) perfectly rational agents select the worst equilibrium, while all equilibria are trembling-hand perfect and proper. With the possibility of forming an institution, all these concepts predict the Pareto-optimal equilibrium as the unique outcome. As we do not observe this outcome in our experimental results, only the Agent QRE model with bounded rationality can explain our data.