It's hard not to be a little mystified by the long-awaited stress-test results. A few weeks ago we worried that several big banks might be insolvent; yesterday the government told us that the 19 stress-tested banks collectively need a meager $75 billion (a substantial portion of which they can generate simply by converting the preferred shares the government currently owns to common stock). One struggles to make sense of it.

But if you actually read the government's stress-test report, you come away thinking the numbers are defensible. (I'm not saying I necessarily agree with them, but you see a legitimate rationale). In fact, the numbers look pretty similar to some recent IMF numbers that are widely viewed as credible. The IMF expects the entire U.S. banking system to lose about $620 billion this year and next year; it puts the cumulative losses from 2007 (the beginning of the housing bust) to 2010 at $1.2 trillion. For its part, the stress-test scenario anticipated about $600 billion in losses for the 19 banks this year and next year, and $950 billion in cumulative losses between mid-2007 and 2010. (When you further consider that the 19 stress-test subjects represent about two-thirds of the U.S. banking system, the government's numbers are actually more pessmistic than the IMF's.)

You can go even further in this vein. For example, when Brookings' Douglas Elliott took the IMF numbers, made reasonable assumptions about the banks' future profitability, and required the banks to hold roughly the same ratio of capital to overall assets that the stress tests required, he found they needed to come up with around $200 billion in capital (through some combination of conversions and fresh money). The corresponding number from the strress-test results: $185 billion.

So how do you reconcile that figure with the $75 billion figure you've read about in just about every news account? According to the stress-test report:

The $185  billion  estimated  additional  capital  buffers  correspond  to  the  estimate  that  would  have  applied  at  the  end  of  2008.   But  a  number  of  these  firms  have  either  completed  or  contracted  for  asset  sales  or  restructured  existing  capital  instruments  since  the  end  of  2008  in  ways  that  increased  their  Tier  1  Common  capital.   These  actions  substantially  reduced  the  final  SCAP  buffer.   In  addition,  the  pre-provision  net  revenues  of  many  of  the  firms  exceeded what  was  assumed  in  the  more  adverse  scenario  by  almost  $20B,  allowing  them  to  build  their  capital  bases.   The  effects  of  these  transactions and  revenues rendered  the  additional  capital  needed  to  establish  the  SCAP  buffer  equal  to  $75  billion.    

Translation: A number of banks recently raised money by selling off subsidiaries (or are about to), or they converted preferred stock into common stock, and their first-quarter profits came in a lot higher than expected (about $20 billion higher). If you add up all that new money, it comes to about $110 billion, which shrinks the capital hole from $185 billion to $75 billion.

As I say, it all seems defensible to me.

For what it's worth, my concerns are as follows:

1.) Elliott's $200(-ish) billion figure is based on a forecast of what the IMF expects to happen. The government's comparable $185 billion figure is based on something it doesn't expect to happen; it takes great pains to insist that the stress-test scenario is much more pessimistic than what it actually expects. This leads the government to argue that it's acting from an abundance of caution rather than based on what the banks realitstically need.

Now, it may not matter: $X-billion in capital is $X-billion in capital, whether you raise it because you absolutely need it or because you're being extra-cautious. But my concern with the "abundance of caution" mindset is that it could lead you to undershoot, which may explain how we moved so abruptly from $185 billion to $75 billion. (If you think you're going above-and-beyond, why not go slightly less above-and-beyond and give the banks credit for have raised more money than you expected). And I don't think this is a moment to be undershooting.

2.) If I were a bank supervisor (the anonymous government officials who carried out the stress tests), and I saw first-quarter bank profits come in $20 billion higher than I expected--and at a time when banks were chafing against government intervention and had an enormous incentive to show they could go it alone--I guess I'd be more inclined to question that $20 billion figure than to question my original expectations. All the more so when I started to learn about the various accounting tricks the banks had used to generate it.

3.) There are big information asymmetries involved in this exercise: The government has just combed through the books of 19 banks; it knows what their books of loans look like, what their portfolios of investments look like. You and I have no such information available to us. That makes it impossible to verify this work on our own.

To give you one example: I've been tooling around the MetLife annual report lately. And one of the things you notice if you read it is that the company (which is now officially a bank) has just under $30 billion in what it calls "gross unrealized losses" on various debt securities (like mortgage-backed securities). These are basically declines in value that MetLife doesn't declare as losses because it plans to hang on to them for a while and it considers the declines temporary. Or, as the annual report elegantly puts it: "Based upon the Company’s current evaluation of these securities in accordance with its impairment policy, the cause of the decline being primarily attributable to ... an extensive widening of credit spreads which resulted from a lack of market liquidity and a short-term market dislocation versus a long-term deterioration in credit quality ... the Company has concluded that these securities are not other-than-temporarily impaired." 

Now, I worry that a lot of these securities are actually permanently impaired.* It certainly wouldn't shock me if half or two-thirds of that $30 billion is a lost cause. But when I look at the stress-test results for MetLife, I see that the government assumes no more than $8.3 billion in losses this year and next year. (I say "no more than" because the category in the stress-test report is broader than the debt securities I'm looking at in the MetLife annual report.)

For all I know, the government is exactly right. As I say, the regulators know precisely what kind of securities MetLife has, and they're in a much better position to assess what sort of losses MetLife could suffer under tougher economic conditions. But given my anxieties, and given the importance of the determination, I'd feel a lot better if we could open this stuff up and get a third, fourth, fifth opinion on it. We obviously can't do that, given the sensitivity of the information. So I'll just have to deal with sleeping about 10 percent worse at night.

--Noam Scheiber

*Given that I'm talking about the finances of a specific bank, I should disclose that some of TNR's investors may own positions in certain banks (though I couldn't tell you more than that--that's all I know).