This post from Mark Thoma is useful in spelling out some of the accounting behind the Geithner plan and its old incarnation due to Paulson and co.  But we cannot asssess the policy unless we come to grips with the Treasury’s motives for intervening in the first place.  When we do the picture changes a lot and it becomes clear that this amounts to a blanket insurance policy for the banks.

Suppose that a bank has a stockpile of toxic assets, and suppose that this bank is solvent only if those assets value at least $X.  When TALF comes to negotiate the purchase of these assets, we know that the bank will not accept anything less than $X for them.  Accepting less than $X turns a concern which is potentially solvent (under rosy assumptions about a recovery in the market for the assets) into one which is certainly insolvent.  The balance sheet woud now be transparent and the bank will be shut down.

So TALF either results in no sale, or a sale above $X. A sale at $X or higher ensures that the bank is solvent and therefore amounts to guarantee of the bank’s liabilities.

I am not expert enough to know whether guaranteeing the bank’s liabilities is a good idea (I suspect it is not the best), but I can say this.  If free insurance is what the Treasury wants out of TALF, then TALF is a bad way to do it.  A simpler and far better way is to simply declare that the bank’s liabilities are backed by the government.  It amounts to the same thing if TALF were to work properly.  But there are many ways TALF could go wrong.

For example, there is no assurance that under TALF the bank will actually use the $X cash from the sale to stay in business.  No doubt Geithner will make sure that an AIG-style transfer to executives and shareholders will not happen but there are too many other possibilities to guard against in law.  By contrast, a real insurance guarantee means that the money does not change hands until the creditors come calling and then it goes directly to the creditors without the bank ever touching it.

A second problem with TALF is that the government typically does not know the exact value of $X.  To be sure that it actually covers $X, it would have to accept the high probability that it overpays.  With a real insurance policy there is no need to guess at X because it will be revealed when the bank defaults.

BTW, I made a related, but somewhat different point about TALF’s predecessor here (pretty technical.)