I don't understand the whole "too big to fail" idea. They say that the collapse of a large institution would have a domino effect through the rest of the banking world, but how does it ever get that bad to begin with? If a particular institution is getting too big, shouldn't the other players want to hedge against its failure to protect themselves against the low-probability but high-cost situation in which a really big bank fails? Is this a situation where the market would behave like that in the ideal case (with perfect information, instantaneous communication, unbounded computation, etc) but we end up with "too big to fail" banks due to some sort of inefficiencies that occur in practice? Or is this situation due to government meddling (bailouts, guarantees, etc)? Or are these swings just due to humans acting emotionally/irrationally (e.g. in response to scare mongering news stories)?

This is just a total guess, but to me it seems like it gets this bad partly because of a feedback cycle between inefficiencies of real-world markets and government meddling. The players in the real world markets don't have perfect information and instantaneous communication, so they end up not balancing their potential payoffs against the risks. The government (or central bank, whatever) steps in to rescue the system when the situation gets bad by propping up failing banks. Sounds good so far, but the problem is that this effectively functions like a zero-cost insurance policy and insulates the banks from some of the true risk of their investments. This encourages the banks to take riskier investments, since they get part of the risk paid for for free. Now you've got a positive feedback loop: banks take riskier investments due to gov't guarantee, gov't steps in when the investment goes bad, banks gain confidence in gov't bailout and make more even worse investments, rinse & repeat.

It also seems like the governmental "insurance policy" encourages the kind of interdependency where one bank's failure can bring down other banks. It's more likely that the banks can collect on the insurance policy if many of them go down at once, but if only one bank is in trouble it's more likely that the government will just allow that one bank to fail. (I'm not proposing a conspiracy theory; I'm just saying that this "insurance" shifts the banks' costs/benefits towards mutual dependence and reduces the attractiveness of pursuing diverse/unconventional strategies.)

OK, that's my half baked hypothesis -- what's the real explanation?