We study the macroeconomic effects of bank runs in a neoclassical growth model with a fully microfounded banking system. In every period, the banks provide insurance against idiosyncratic liquidity shocks, but the possibility of sunspot-driven bank runs distorts the equilibrium allocation. In the quantitative exercise, we find that the banks, for low values of the probability of the sunspot, choose a contract that is not run-proof, and satisfy an “equal service constraint” if the run is realized. In equilibrium, a shock to the probability of a bank run leads to a drop in GDP of between 0.001 and 5.6 percentage points.