The idea that any financial institution that's "too big to fail" (TBTF) shouldn't be allowed to exist has evidently gained significant support. What's discouraging is that people seem to think this is a sophisticated and workable strategy that we should actually pursue, mostly because people like former IMF official Simon Johnson say things like this:
I have yet to hear a single responsible official in any industrial country state what is obvious to most technocrats who are not currently officials: anything too big to fail is too big to exist.
This is the kind of thing people say at cocktail parties to make themselves sound smart without having to do any serious work. It's a preposterous idea.
There's a large literature on TBTF (see e.g., here, here, here, and here, but Johnson is clearly not familiar with any of it. I've followed this debate for quite a while, and I've never seen any expert in this field propose a policy of "too big to fail, too big to exist," so the idea that enacting such a policy is "obvious to most technocrats" is simply not credible. But as Dani Rodrik put it:
The [IMF] is staffed by a large number of smart economists, who lack much connection to (and appreciation for) the institutional realities of the countries on which they work. Their professional expertise is validated by the quality of their advanced degrees, rather than by their achievements in practical policymaking. This breeds arrogance and a sense of smug superiority over their counterparts -- policymakers who must balance multiple, complicated agendas.
Setting a limit on bank size is obviously not feasible. How do you think the FDIC resolves most failed banks? It uses Purchase and Assumption (P&A) transactions, in which another bank assumes certain liabilities of the failed bank in exchange for some or all of the failed bank's assets. The only reason the WaMu failure went so smoothly because JP Morgan agreed to steal buy WaMu through a P&A. Say the limit on bank size is set at $100 billion in assets. If a bank with $99 billion in assets failed, the FDIC wouldn't be able to use almost any form of P&A--including a Whole Bank P&A, which is the the most common resolution method--because the acquiring bank would be over the $100 billion limit after the acquisition. The fact that there are larger banks that can gobble up failed small banks through P&As is one of the biggest reasons the FDIC has been so successful in resolving failed banks.
In any event, it's clear that the people pushing the "too big to fail, too big to exist" policy fundamentally misunderstand the nature of the TBTF problem. The issue isn't the size of a financial institution per se, but rather its systematic importance. As Alan Greenspan famously framed it:
[T]he issue is an organization that is very large is not too big to fail, it may be too big to allow to implode quickly.
The reason this distinction is so important is that defining "too systematically important to fail" ex ante is simply not possible.
To be sure, systematically important financial institutions are usually quite large, but they don't have to be. Examples of large financial institutions that weren't systematically important include Drexel Burnham and BCCI, both of which failed without causing systemic damage.
Systematic importance has to do with the nature of the role an institution plays in the financial markets. For example, the Depository Trust & Clearing Corporation (DTCC) clears and settles virtually all securities transactions in the US--in 2008, it settled $1.88 quadrillion in securities transactions. If the DTCC failed, financial markets would literally cease to function, as clearing and settlement are necessary to complete a securities trade. It would be Mad Max. What makes the DTCC "too systematically important to fail" isn't its size though, it's the critical role it plays in the financial markets, and the fact that no firm in the world could immediately step in and start settling $7.5 trillion of securities transactions every day if somehow, God forbid, DTCC ever failed.
Similarly, only two institutions provide full clearing and settlement of government securities : Bank of New York and JP Morgan. Both of these institutions would be large enough to qualify as "systematically important" even if they didn't clear and settle Treasuries, but the point is that they would also qualify as systematically important even if they were relatively small. As long as they were one of the two institutions that cleared and settled Treasuries, they would be considered systematically important.
Outside of the payment system, there are several key markets where a material disruption could set off a dangerous chain reaction. The interbank market is one obvious example. A bank would be considered systematically important if it accounts for a large enough share of interbank lending. What constitutes "a large enough share of interbank lending"? No one really knows, but any definition of "too systematically important to fail" would have to identify that threshold. The short-term repo market is another key market where the failure of an important player could have systemic consequences, because so many financial institutions fund their operations through repos. Most experts consider the largest derivatives markets to be "key markets" as well, because the failure of an important player in one of these markets could force counterparties to liquidate the collateral posted against these derivative contracts. If the collateral the failed institution had posted against its derivatives was sufficiently similar, such forced liquidations would depress the value of the assets, and would impose substantial losses on any institution that held a significant amount of these assets. These key markets can be thought of as the pressure points of the international financial system.
The point is that there are a number of complex, interrelated factors that make a financial institution "systematically important." Size is one factor, but it's hardly determinative. Other factors include an institution's relative importance to key markets, the similarity of exposures between an institution and its counterparties, how highly leveraged an institution's counterparties are, etc.
So why is it impossible to define "too systematically important to fail" ex ante? Because these factors are constantly in flux. The key markets--the ones where the failure of an important player could have severe knock-on effects--are especially difficult to identify in real time, because the composition of the major participants in a given market can change rapidly. In other words, the pressure points in the international financial system aren't static. Any statutory definition would be hopelessly behind reality.