Now that the TARP bill has passed, Neel Kashkari and his team at the Treasury Department are going to have to very quickly come up with a fair and transparent way of spending $700 billion on a very wide range of complex financial assets. How are they going to do that?

The most detailed public proposal to date comes from Lawrence Ausubel and Peter Cramton at the University of Maryland. In a densely argued 21-page paper, they propose a two-part reverse auction, starting with the more commonly held assets and then moving on to the stuff that is really difficult to value.

The cleverest bit of the Ausubel-Cramton proposal is that it recognizes that banks don't have a single price below which they're unwilling to sell any given asset. Instead, as the price falls, they are still willing to sell some, just not all, of their holdings.

Let's say you're a bank with $100 million, face value, of a subprime-mortgage-backed security. You would love to get that asset off your balance sheet, since the market is worried that if it turns out to be worthless, you might be insolvent. But, at the same time, the bond is still paying out money, and according to your best estimate of its expected future cash flow, if you hold it to maturity you should get a present value of 60 cents on the dollar.

Right now, you can't mark the bond to market, because there is no market for such instruments. The only people remotely interested in buying up subprime-mortgage bonds are the distressed-debt hedge funds, and they're looking to make annualized returns of more than 30 percent.

Enter the Treasury. Kashkari puts you into a virtual room with a handful of other financial institutions that hold that same bond: All of you own $1 billion, face value, of it. He announces at the end of the auction that he's going to buy half of the bonds in the room. And then the reverse auction starts.

The bidding starts high: at, say, 100 cents on the dollar. At that point, you're more than willing to tender all your bonds to be sold—and so is everybody else. So you bid 100 percent of your bonds at 100 cents on the dollar, everybody else does too, and the total quantity of bonds tendered comes to the full $1 billion. Supply ($1 billion) exceeds demand ($500 million). And so the price gets brought down a notch.

At 90 cents on the dollar, everybody's still interested in selling all their bonds. And the same thing happens at 80 cents. But at 70 cents, something interesting happens: Not all the bonds are tendered. You're still happy to sell your full $100 million holding at that point, but when the bids are tallied up, it turns out that only $950 million of bonds have been tendered. Some bank has started paring back its bid.

At 60 cents, you still bid the full $100 million. You think that the bonds are probably worth that much, but you'd much rather have $60 million in nice liquid cash than $100 million face value of bonds you think are probably worth 60 cents on the dollar. Most of your fellow bidders seem to think the same way: The amount tendered at 60 cents on the dollar is $850 million. But it's coming down.

And so the auction moves to 50 cents. Now what do you do? You think that the bonds are worth 60 cents, but the stock market doesn't necessarily believe you, and it would start bidding up your stock price if you said you'd managed to sell off most of your subprime debt at 50 cents on the dollar. On the other hand, you think that the bonds are actually a reasonably good investment at that level. So you tender $80 million of your bonds. That way, you get $40 million in cash—which would certainly be welcome—and you also retain a small $20 million holding in a security which you think will bring in a good 60 cents on the dollar if you hold it to maturity. The holding is small enough that the market no longer worries about your solvency, and everybody comes away happy.

At 50 cents, however, there's still $650 million of bonds tendered. And so the price comes down again, to 40 cents. At that point, you really don't want to sell—you think that there's a lot of upside in the security at that level. But you also need the liquidity, and you want to be able to demonstrate that you've managed to get at least some of your subprime assets off your balance sheet. So you offer to sell Treasury $40 million of your bonds at 40 cents on the dollar. That brings you $16 million in useful cash and reduces your subprime exposure by a significant amount. You don't like it, but it's better than selling nothing at all and ending up where you started.

At 40 cents, the total number of bonds tendered is the magic $500 million, and the auction clears. Some banks will have ended up tendering all of their bonds at that level—maybe they're more pessimistic about that security's intrinsic value, or maybe they just really need the money or to get the bonds off their balance sheet. In any event, there's now a public and transparent price for the bonds: 40 cents. That can be used to mark all the bonds that were not tendered to market, rather than guessing wildly on the basis of credit-default swap prices or some model.

This is the basis of the Ausubel-Cramton approach. It's quite elegant, in that clearly solvent banks will tend to drop out of the bidding quite early, once it reaches their own estimation of the value of the securities, while more troubled banks will end up selling bonds to Treasury at discount prices. Taxpayers are protected, and although the more troubled banks will end up with more government funds, they'll still be losing out on an intrinsic-value basis.

Once marks are set for a wide range of relatively widely held subprime and other mortgage securities, the Ausubel and Cramton proposal foresees a second round of auctions for less liquid assets, which involves using the prices generated in the first round as inputs to help get at least a rough idea of how much the more complex instruments might be worth. Conceivably, there could even be a third round along similar lines. Eventually, the $700 billion is spent; the government ends up with a large number of mortgage-backed assets, as well as various chunks of bank equity that were extracted along the way from the largest participants in the program.

There is a problem with this model, however.