The Daimond-Dybvig model is a piece of garbage. While part of me is happy to see U of C get another Nobel, I really don't think Daimond deserves it. DD is, in fact, a strawman argument meant to justify the policy of FDIC insurance. It does not realistically represent real banks for a whole range of reasons:
1) Real banks have capital cushions, i.e. investors, who provide extra security to the bank. Daimond-Dybvig assumes the bank's deposits can only be covered by other deposits. The claims against the bank in DD's model are a hybrid of debt and equity the likes of which we do not see in the real world. In real world banks, equity and creditors (depositors) are two distinct classes that receive different treatment—and equity holders receive a larger potential payoff in exchange for a higher risk of loss. This arrangement shields depositors to some extent.
2) Similarly, it ignores the possibility of an inter-bank loan to cover unexpected withdrawals.
3) All deposits mature at the same time in their model, but not in real life. The fact that deposits mature at different times and you can be rewarded for keeping your money in the bank for a little longer allows agents to gamble on their place in line. If you build an agent based model of bank runs, as my friend and colleague John Schuler has, you can see that real world behavior looks much different from what is predicted by Daimond's model.
4) In their model, you are rewarded for pulling your money out pre-maturity. In real life, you would likely have an early withdrawal penalty, not an early withdrawal slightly smaller reward. While someone could object that it is the opportunity cost that matters, people looking to withdrawal early would enter into Daimond's contract, but would never enter into one with an early withdrawal penalty. A real world CD contract brings in people who are not planning to touch that money while DD's contract brings in people who might be happy to take that lower payoff.
5) In the real world, even if redemption for cash is suspended, "bank money" (i.e. fiduciary media as Mises would call it) can still be traded. This also reduces the chance of a bank run. In short, you can still write checks and pay off your credit cards with you bank deposits in the real world even if you are temporarily barred from pulling out cash.
6) The model gives the government the power to discount the value of redeemed deposits, let's say allowing people to take out 80 cents on their dollar when processing FDIC insurance claims, but does not allow the bank to do this. By adding this constraint on the bank, they engage in a kind of question begging. Why allow the state to do this, but not allow the bank to build such an emergency measure into their contractual arrangement?
If you aren't familiar with Daimond-Dybvig some of what I said might not make sense. But the upshot of my remarks is this: Their model does not resemble a real world bank in any way.
Not only is Daimond-Dybvig wrong, but it isn't even an interesting wrong model. If we are going to give someone a Nobel for being wrong, I would much rather give it to Nordhaus for his political business cycle theory. There are other wrong theories that helped facilitate discussion and research, and helped push economics forward, better than DD did.