There is no perfect answer to TBIF but the two most promising approaches are as follows:  First, create a resolution authority (or a special bankruptcy court) so that large banks can be closed down in an orderly fashion without excessive disruption to the system as a whole.  The most difficult part of this approach is resolving international banks and that requires cooperation with other countries, especially other financial hubs such as London.  The resolution process should heavily penalize managers and shareholders and make bond holders take losses.  A special fund should be available to make sure the institution is kept operating until it can be sold off or shut down.  This fund should be drawn from a levy on other financial institutions, especially other large financial institutions.

The second approach is that financial institutions should be required to hold more capital the bigger and riskier they get.  The approach should not be punitive but should capture the additional risks imposed by large and complex institutions on the financial system as a whole.  Large banks should also be subject to additional scrutiny from regulators to make sure they are following sound risk management strategies.

The two biggest problems in the financial crisis were, first, that financial institutions did not have adequate risk management rules or did not follow them if they had them.  Second, regulators pored over the books of the banks but never really tested whether they were keeping their risks under control.  These problems were not specifically problems of size, but of poor management and regulatory practices and these must be changed going forward.

There are means other than draconian limits on size to credibly prevent government bailouts of large institutions. And there are large social gains from retaining large, complex, global financial institutions.

As I argue more at length elsewhere, it is worth preserving large financial institutions four reasons. First and foremost, financial institutions need to be large to operate with global scope because their clients are large and global. Small, local banks simply could not provide global corporations the same physical capabilities for trade finance, foreign exchange contracting, and global capital access that large global financial institutions can. 

Second, there are production economies of scope that offer benefits when financial firms combine different products within the same firm. Economies of scope among products implies economies of scale within finance suppliers, since small financial firms cannot afford the overhead that comes from building platforms with many such complex products.

Third, many of the gains of consolidation accrued to customers, not banks, in the form of cheaper and better financial services. Among the many examples, perhaps the greatest accomplishment of global finance in the past two decades has been the replacement of crony banking networks in emerging market countries with branches of large global banks. 

Fourth, global financial institutions also have made stock, bond, and foreign exchange markets globally integrated and more efficient.

We can solve the too-big-to-fail problem without destroying global finance.

I am far more comfortable with a regulatory system that gradually penalizes size, however, through progressive higher capital/liquidity requirements. A graduated system of regulatory obligations would more consistent with market-principles, and let the organizations and the market figure out how to best cope with the higher costs their size (and inter-connectedness) impose on the financial and economic system as a whole.

Furthermore, to the extent it does not generate excessive costs of its own, large troubled financial institutions that are being resolved by the authorities (under needed new legislation, giving regulator bank-like resolution powers) should b dismembered so that they don’t come back to haunt us again. But regulators also should have the option not to conduct surgeries if they are significantly more expensive than what they would otherwise do to meet a “least cost resolution” standard.