I’m a big advocate of much strengthened financial regulation. One argument I don’t buy, however, is that we should try to shrink financial institutions down to the point where nobody is too big to fail. Basically, it’s just not possible.I also think that the "shrink the banks" philosophy rests on some shaky assumptions about the nature of financial crises. The idea is that crises begin when the failure of a large firm leads to chaos in the system. In retrospect, this almost always seems like the correct story: as long as we have a core of large, important banks, any destructive financial panic is likely to consume at least one of them, and ex post facto we can finger the first such failure as the point at which the system really started to fall apart. This hardly means that we can avoid panics by writing large banks out of existence.
The point is that finance is deeply interconnected, so that even a moderately large player can take down the system if it implodes. Remember, it was Lehman — not Citi or B of A — that brought the world to the brink.
And as far as I know, there never was a time when policymakers could have viewed the collapse of a major money center bank with equanimity.
To claim that a system where risk is distributed among many small banks will demand less government intervention to prevent crises, I think you need to make a strong argument for the benefits of diversification. Replacing one large institution with twenty small ones does very little if the twenty new banks fail in exactly the same way. You need to assume that this is unlikely to happen—that the individual risks from each smaller bank will aggregate to a significantly less risky whole.
I think this is a bad assumption. Recall that the most spectacular mispricing of risk in our current crisis arose in exactly the same way—statistical magic transformed baskets of extraordinarily risky securities into assets "riskless" enough to be rated AAA. This worked for a while, but ultimately the logic of crises took hold. In moments of systemic danger, you can't count on risks aggregating away. All the correlations run to one, and it's no safer to have a crowd of small banks than a few large ones.