I really disliked Tim Geithner's 'stress test'. Taking over companies that could be insolvent in the future is a horrible precedent, setting a vague standard for solvency, because if you are insolvent in the future under the government's opinion, you are insolvent today. Thus, insolvency is not the exit criteria, rather one has to plan for future 'future' insolvency as defined by the government. This is a political football, meaning, it's all lobbying and appeasing regulators to maintain market power and beat down new entrants. I noted in his WSJ editorial, Geithner mentioned that people will be more confident in banks after having passed this vetting process, and that may have been the motivation, but passing a politically infused stress test is not reassuring.
On the other hand, his new plan is to basically match fund private investors in buying distressed securities, with over $500B set aside. This mitigates poor incentives, because they are not providing the first loss or senior piece, just sharing parri-passu. They are not giving purchasers leverage, even. An investor who thinks it is a bad risk-return investment will not be incented to invest, because this government involvement does not change the payoff space. This is just adding liquidity, and is not a subsidy, because the expected value of this is zero if we assume market prices are expected values of future payoffs. The cost is actually negative if we think the markets are in a panic, but costs are quite high if we think this is the mid point in a massive financial cataclysm.
Thus, Krugman thinks this is a huge subsidy because he thinks market prices are way too high for CMBS and RMBS securities, though I think this is really because he believes anything that would justify greater government control of economic activity. He has presented absolutely zero data on the prices and corresponding historical loss curves for the various vintage-product types (eg, Alt-A mortgages, 2006, average current price and losses). He doesn't have the data, he wouldn't know how to set it up. It's the kind of detail he is ignorant about but does not think is important. Details matter. Subprime, Alt-A, conforming,vintage, seasoning all matter. AAA 2007 subprime tranches trades at around 25% via the Markit ABX index. Is that really reasonable? What are the losses on the collateral? He just wants to take over the banks asap because asymmetries and imperfect information prove markets are inefficient (heh).
I found 94 subprime mortgage tranches, and some referred to the same underlying mortgage pool. You can find these by going to the Markit ABX index page, looking at the constituents, and see they refer to 20 tranches that vary by seniority for the various grades (eg, CWALT 2007-21CB M Mtge, BSABS 2007-AQ1 A1) . Anyway, the average cumulative 90+ delinquency is about 25%. That's pretty high historically, but after seasoning 21 months, this means a conservative estimate of total 90+ delinquency is going to be around 50%, and losses on those delinquencies no more than 50% (housing prices did not fall more than 50% in most places, on average). Thats for all of subprime in 2007. The market price for AAA rated assets (better than average) is a mere 25 cents on the dollar, suggesting the average subprime mortgage is priced below that, say 10 cents on the dollar. This is way too low.
I suspect most banks will be unwilling to sell off assets at market prices, because the market prices are so low. In that case the plan 'fails', but it would also then cost us nothing, and in the meantime not screw anything up (in contrast to the stress test).