Reducing Discretion in Financial Regulation

A lot of people have concerns about the level of regulatory discretion involved in Chris Dodd’s financial reform proposal. Most of the ideas for reducing discretion that I’ve seen involve introducing hard floors, like the way the House says capital requirements can’t drop below 15:1. But writing in National Affairs, Oliver Hart & Luigi Zingales attempt to develop an approach that reduces discretion while maintaining flexibility. Their idea is that banks should be required to maintain two tiers of capital. They explain:

Once we understand that the issue addressed by “too big to fail” is the interconnectedness of large financial institutions, and therefore the stability of the larger system, we can make some important distinctions. Not all of the debt held by large financial institutions, and not all of the transactions they engage in, are systemically relevant in this way, and so in need of such total protection. Specifically, long-term debt is not “systemically relevant.” There is no reason for a large financial institution to hold bonds or other long-term debt in other financial institutions. This debt mostly resides in the massive portfolios of mutual funds and pension funds, which can absorb losses in the value of such debt in the same way they absorb losses from equity investments. A default on that debt, therefore, would not trigger a cascade of bank failures the way a default on short-term debt could.

With that in mind, a bank would hold two tiers of capital. One, a “senior” tier tied to the systemically important obligations. The second, a “junior” tier that covers longer-term obligations. Because the junior tier would be non-guaranteed, there would be a trade in credit default swaps on the junior capital. Regulators could then set a threshold such that when CDS prices on a given institution hit a certain level, it automatically triggers a stress test:

If the trigger were to be set off by a too-high CDS price, the regulator would be required to carry out a “stress test” on the financial institution to determine if it is indeed at risk. In a stress test, regulators use sophisticated algorithms to run “what if” scenarios that examine whether a financial institution has sufficient assets to survive serious financial shocks. A stress test should precede any other action, so that extraneous panic is not allowed to bring down financial institutions unnecessarily. If, for instance, a few significant hedge funds or other investors lost confidence in a bank on the basis of a rumor or misperception about its strength, and began to buy credit default swaps as protection against its failure, the CDS price would rise and might trigger regulatory action. It is important that the regulator first test the validity of the concern before acting on it.


If the bank passed the test and showed the regulator that the CDS price was not accurate, the regulator would then declare the company adequately capitalized. But if the bank failed the test, the debt was found to be at risk, and issuing equity did not improve its situation, the regulator would replace the institution’s CEO with a receiver or trustee. This person would be required to recapitalize and sell the company, guaranteeing in the process that shareholders were wiped out and creditors — while not wiped out — received a “haircut,” meaning that the value of what they were owed would be reduced by some set percentage. That haircut is crucial in ensuring that the market prices credit default swaps in a way that takes regulator interventions seriously — showing that creditors will pay a price for an institution’s failure — and so makes the trigger more reliable.

To an extent this merely pushes the discretion part of the story down to the “stress test” level, rather than actually eliminate it. In this sense, I think Hart and Zingales are sort of exaggerating the extent to which their proposal differs from the administration’s white paper or Chris Dodd’s bill. It is different, but it doesn’t really solve the discretion/time-consistency issues nor am I sure that those issues really can be solved.