And Mark Thoma also links to a similar question from Joseph Stiglitz: are the banks too big to restructure?
First an answer to Robert's commenter:
The most obvious difference between a bank or building society and a normal business is that the government guarantees the deposits of a bank. If this didn't happen, the depositors would be creditors like any other. But the other distinction, which I'll come to in a few paragraphs, is that banks are much more closely embedded in the operation of the rest of the economy than most other companies.
If they were to be treated as a normal business, then the liquidators would come in and start to sell off their mortgage book (probably at a discount) and the creditors (bondholders, depositors and suppliers) would have to queue up to get back whatever they could from the proceeds. If there was anything left the shareholders would get it - unlikely since the current illiquidity of the market for mortgages means there would probably be nothing left.
But because banks are regulated and the government insures the deposits, they reserve the right to step in and immediately eliminate the shareholders. They pay depositors back immediately and then make a decision about how gradually to sell off the mortgages. In theory they might decide not to sell them off at all, but to hold onto them until the homeowners pay them back over 25 years as normal (some mortgages of course will be partly paid already and will finish paying off sooner than that).
So the question then becomes: how are the bondholders treated? In a normal liquidation, they would not get their money back: instead they would take control of the assets and try to get whatever they could for them. But in a complex bank liquidation this may end up costing so much in legal and accountancy fees that they'd end up getting very little back. Plus, the government - having paid off the depositors - is first in line. So what then (this is where we get onto the Mark Thoma/Joseph Stiglitz link too)?
In fact, the government has mostly provided enough money to the big banks to stop them from going into liquidation and to ensure that the bondholders don't lose out.
In a rational market, making bondholders absorb a share of losses - instead of the taxpayer filling the gap - would not affect future lending by a single cent. Everyone would dust themselves off, write off the old losses and start lending again (at least those who still have anything to lend).
Or maybe the problem will not be a refusal to lend, but simply that they'll demand interest rates which make the business models of banks unworkable. And simultaneously, that they'll enmesh the banks in vast lawsuits which would make them unmanageable.
Which brings us to the other half of the rationale: unlike most firms, banks are so critical to the operation of the economy that if their business model were to disappear, the rest of the economy would be thrown into chaos while restructuring takes place and the financial sector finds a new way to operate.
So we agree to keep them going for now - perhaps while a new business model is sought, or perhaps forever. It does amount to a public subsidy, but maybe the banks have such huge externalities that the subsidy is economically justified.
An effective financial sector (and despite popular belief, we do have one that's effective in most ways) does vastly reduce transaction costs for the non-financial parts of the economy. With the economy operating as it now does, those transaction costs could easily represent 5-10% of GDP without a banking sector, and the economy would immediately become highly distorted as people tried to work around them. To prevent an instant depression, the banks are worth preserving.
It's right that shareholders at least are mostly wiped out, as they have been - because they're the easiest class of asset to deal with.
But on both counts, government should be thinking very carefully. Are the bondholders as irrationally loss-averse as the current policy implies? Even if they are, then lending might still continue if the finance sector can find new ways to route money from wherever it ends up after a restructuring, back into the bond markets again.
So the argument for rescuing banks actually depends on both arguments together: are bondholders loss averse, and is the finance sector too inefficient to quickly restructure and find new sources of capital to lend? If both are true, then government or central bank rescues are justified.
And if the rescues result in restricting the activities or leverage of the financial sector, then this will prevent it from having exactly that flexibility. Perhaps this is justified for other reasons, but it does run the risk of making the rescues self-fulfilling. Because of the bailouts, the bailouts are proven to be necessary. On the other hand, bailouts prevent the money supply from contracting as much as it otherwise would, meaning that the central banks need to print less money than they otherwise would. Perhaps that's why all the money the Fed is printing is sitting on its balance sheet as undrawn bank reserves. It's enough to make your head spin.