Imagine an auction in a looking-glass world, with the auctioneer offering cash to the highest bidder and the participants frantically outbidding each other with a jumble of assorted assets. Should a million dollars be sold to the man in the front row for his bundle of 2006-vintage toxic mortgage securities? Or the lady behind him, for her 2005-vintage offering?

That was the auction the US Treasury was hoping to hold until it abruptly changed its plans in November. It was going to spend up to $700bn – the “Tarp” (toxic asset relief programme) fund – buying a variety of toxic assets from banks. The idea made sense: by establishing a market price for these dubious assets, the auction would have improved transparency and helped solvent banks to prove that they really were solvent.

The Tarp fund was then cannibalised to recapitalise banks (probably a good idea) and then dole out suitcases of cash to all-comers (a less good idea). But the original auction concept should be resurrected, because it solves a problem that has not gone away.

And yet – how could the looking-glass auction work without the US Treasury overpaying for junk? Holding a single auction with a single price would be a disaster: only the most worthless assets would have been offered for sale. To do the job properly and establish realistic prices, the auction needs to distinguish between different assets: some good, some bad and some ugly.

But simply holding many different auctions is not much better. Once finished, a bank might be surprised by the prices, wishing it had sold more of one kind of asset at a generous price, and fewer of its others. Not only would the banks have acted differently with hindsight, but the Treasury would want them to, in the interests of higher revenue and more price transparency.

So two economists from the University of Maryland, Larry Ausubel and Peter Cramton, proposed a dynamic design that would have allowed banks to adjust their bids as the auction proceeded, shifting their emphasis to compete more aggressively wherever prices seemed tempting. The auction was tested to destruction by graduate students and seemed to work well. But there is a flaw: each auction would take a day, with the auction prices affecting financial markets and the markets affecting the auction prices.

However, a solution was already being developed to answer a similar problem for the Bank of England. Paul Klemperer, one of the economists behind the 3G mobile spectrum auctions in the UK, has published a paper explaining how the Bank might auction off loans secured against different qualities of collateral.

He suggests having banks simultaneously submit combinations of bids. Each bank would be considering different scenarios – one in which loan rates were high and it preferred to borrow the absolute minimum; another in which rates were low and it happily offloaded collateral to the Bank of England.

The same approach could work for a Tarp auction, too; Klemperer and three economists from Stanford have been working out the details. A computer would compile bids from both sides, with both the banks and the US Treasury saying in advance what they would be willing to buy or sell at different possible prices. The computer would calculate the result. Because each bidder submitted bids to cover each eventuality, the auction should be efficient and nobody would regret having told the computer the truth.

Whether the US Treasury will relent and return to an auction remains to be seen. The Tarp auction is a fiendishly difficult design problem, but it looks solvable. At the time of writing, it seems that the Treasury prefers to spend the cash ad hoc. Shame.