An experimental exploration of self-fulfilling banking panics : their occurrence, persistence and prevention

This paper tests the possibility and the degree of persistence of self-fulfilling banking panics through an experimental protocol. It confirms the possibility of pure-panic bank runs as first formalized by Diamond and Dybvig (1983) where all participants panic. However, this situation remains rare as Temzelides (1997) found out in an evolutionary game framework. But self-fulfilling bank runs, with a marked loss of confidence between depositors, are the most frequent cases in the experiment. This supports the idea of strong coordination failures as observed in the experimental literature on coordination games. Panics are proved to be persistent phenomena which are difficult to prevent. However, it seems to be possible to curb them through a learning effect caused by a temporary but long enough suspension of the deposit availability, combined to a "narrow banking solution" which makes banks more liquid. Lastly, panic prevention requires a full deposit coverage to be effective. Even a 75% deposit coverage rate is not sufficient, while lowering the coverage rate to 25% leads to more severe bank runs. This suggests that the moral hazard issue should not be tackled through a lower deposit coverage, especially in emerging countries' banking systems where depositors are likely to lose confidence