Experimental evidence of bank runs as pure coordination failures

We investigate how coordination requirement, measured by the coordination parameter, affects the occurrence of miscoordination-based bank runs in controlled laboratory environments. We identify an indeterminacy region of the coordination parameter such that games with the parameter within the region have varying coordination outcomes and exhibit persistent path dependence. Experimental economies with the parameter above (below) the region stay close or converge to the run (non-run) equilibrium. Switches between the two equilibria occur even with fixed economic fundamentals. The experimental results are well accounted for by a version of the evolutionary algorithm that uses experimentation rates estimated from the experimental data.