The Boston Fed has released a working paper by Michael Manz, The Optimal Level of Deposit Insurance Coverage, in which a model of depositor behaviour leads to some interesting conclusions:
The model derived in this paper allows for a rigorous analysis of partial deposit insurance. The benefits of insurance involve eliminating inefficient withdrawals and bank runs due to noisy information and coordination failures, whereas the drawbacks consist of suppressing efficient withdrawals and of inducing excessive risk taking. A hitherto hardly noticed conclusion is that a high level of coverage can even be detrimental if bank risk is exogenous, because it undermines the occurrence of efficient bank runs. An extended version of the model shows that systemic risk calls for a higher level of deposit insurance, albeit only for systemically relevant banks from which contagion emanates, and not for the institutions that are potentially affected by contagion.
A vital contribution of the model is to provide comparative statics of the optimal level of coverage. In particular, the results imply that while tightening liquidity requirements is a substitute for deposit insurance, increasing transparency is not. Rather, the optimal coverage increases with the quality of the information available to depositors. Perhaps surprisingly, the degree of deposit insurance should not vary with expectations regarding the development of the real sector. This suggests that countries that in the past turned to increased or even unlimited deposit insurance as a reaction to a crisis, such as Japan, Turkey, or the United States, would do well to pause for thought on whether this is the right measure to strengthen their banking systems. The model also demonstrates why the presence of large creditors with uninsured claims calls for a lower level of insurance and why a high coverage is foremost in the interest of bankers and uninsured lenders. Moreover, it is consistent with small banks being particularly active lobbyists in favor of extending deposit insurance.
Another key advantage of the model is its applicability to various policy issues. In practice, only a small, albeit growing, number of countries maintaining deposit insurance require bank customers to coinsure a proportion of their deposits. According to the model, however, an optimal design of protection should build on coinsurance rather than on setting caps on insured deposits. It further indicates that deposit insurance becomes redundant in combination with full bailouts or optimal lending of last resort. While an unconditional bailout policy is about as inefficient as it can get, an optimal LolR policy combined with perfect public disclosure comes closest to the first best outcome in terms of welfare. Yet such an optimal policy, which requires protection to be contingent on bank solvency, seems far more demanding and hence less realistic in practice than unconditional deposit insurance. If regulators or central banks cannot precisely assess whether a bank is solvent, interventions are likely to be a mixture of the benchmark policies considered. Investigating these cases opens an interesting avenue for future research.