Commenter roidubouloi, our resident bank hawk, raises a great point:
It is very difficult to reconcile the IMF figures with the results of the stress tests -- impossible actually. The stated capital of the banking system is on the order of $1 trillion. If there are losses on the order of $1.5 trillion, then the system is insolvent. But if none of the major players is insolvent, merely stretched, where did the losses go?
That gets even harder to answer when you consider that the 19 banks being stress-tested account for about two-thirds of the assets in the banking system. It's tough to imagine the losses being concentrated in that other third, especially since most small banks haven't been placing wild bets on mortgage-backed securities.
But, in fact, its is possible to reconcile the IMF estimates of losses with the likely results of the stress tests, which are expected to require the 19 banks to raise $100-$200 billion in new capital (probably through some combination of new preferred stock, conversions of preferred to common stock, and new issuance of common stock).
Douglas Elliott of Brookings has a really helpful paper doing just that (via Matt Yglesias). The gist is that, while the IMF puts the expected cumulative losses in the U.S. banking system at about $1.17 trillion (cumulative between 2007 and 2010), about $510 billion of these losses had been written down as of the end of last year. If you further assume (not unreasonably) that the banks will collectively earn about $300 billion between now and the end of next year, you're left with about $320 billion in losses that need to be absorbed. Take two-thirds of that to get to our universe of stress-tested banks, and you're at about $215 billion--right at the edge of our range.
From there, the determination about whether the 19 banks collectively need more or less than $215 billion depends on how much you trust our underlying assumptions and whether you think the banks currently have more or less capital than they'd need under normal economic circumstances (i.e., where you want them to end up once the losses are absorbed and the economy recovers). For example, you may think the IMF is way too optimistic and buy the more dire analysis of someone like Nouriel Roubini (who estimates losses to be about 50 percent higher than the IMF), in which case you'd obviously think banks need much more capital. But the $200 billion figure seems reasonable if we're basically living in an IMF world.