We study the effects of deposit insurance and observability of previous actions on the emergence of bank runs by means of a controlled laboratory experiment. We consider three depositors in the line of a bank, who decide between withdrawing or keeping their money deposited. We have three treatments with different levels of deposit insurance which reflect the losses a depositor may incur in the case of a bank run. We find that different levels of deposit insurance and the possibility of observing other depositors’ actions affect the likelihood of bank runs. When decisions are not observable, higher levels of deposit insurance decrease the probability of bank runs. When decisions are observable, this need not to be the case. These results suggest that (i) observability might be considered as a partial substitute of deposit insurance and (ii) the optimal deposit insurance should take into account the degree of observability.