For instructors

Teaching notes

A bank funds long-term loans with deposits that can be withdrawn on demand. A money market fund promises daily liquidity on assets that take time to sell. Each arrangement works as long as most claimants do not ask for their money at once, and each can fail because claimants fear that others will ask.

What the game shows

The game tests whether a group remains in the good equilibrium of a coordination problem in which the bad one is safer for the individual. It shows that a sound bank can fail through expectations alone, that the fear of a run is self-fulfilling, and that a failure in one round shapes behavior in the next. The game also gives a setting in which to discuss deposit insurance and the suspension of withdrawals.

The theory

In the model of Diamond and Dybvig (1983), a bank improves on what depositors could achieve alone by pooling the risk that some of them will need their money early. The same contract creates two equilibria, which differ only in what depositors expect of one another. Where they expect the others to wait, waiting pays more than withdrawing. Where they expect the others to withdraw, the bank must liquidate its assets at a loss, little will be left for those who wait, and withdrawing is the best response. Nothing about the bank's assets needs to change for the second outcome to occur.

In the game here a run is possible only when an early withdrawal pays more than the deposit. The dashboard states how many early withdrawals the bank can absorb before waiting ceases to pay. With the default settings, a bank of five with deposits of 100, an early payment of 120 and a growth factor of 2, waiting remains the better choice as long as no more than three of the five withdraw.

What to expect

Runs are not the usual outcome when coordination is the only difficulty. Garratt and Keister (2009) found them to be rare when the demand for early withdrawals was known and frequent when some depositors were forced to withdraw at random, so that the others could not tell a run from an ordinary need for cash. Arifovic, Jiang and Xu (2013) vary the share of depositors who must wait for waiting to pay. Their groups stayed away from runs when that share was one half or less, ran when it was four fifths or more, and in the range between went either way depending on their own history.

Once a run has occurred it tends to persist. Madiès (2006) reports that panics were difficult to stop, that most runs were partial, and that partial deposit insurance did not prevent them where full coverage did. Schotter and Yorulmazer (2009) study the dynamics and the severity of runs in the laboratory.

Tips and things to look out for

References