Nineteen principles for financial regulation

1.Build a firewall.
The financial system exists to serve the real economy by allocating capital, providing information for other real-sector decisions regarding risk and the timing of saving and dis-saving, providing liquidity for current transactions, and facilitating some government interventions in the economy. But it doesn’t always work right. When it fails, nothing great is lost by its failure per se – but much may be lost as a result of its influence on the real economy. When the financial system functions well, its signals that certain firms should be liquidated are invaluable. But if it starts sending those signals wildly, the cost of liquidating valuable real capital can be devastating. Build a firewall that shuts down those signals when something goes badly wrong.

2.Limit aggregate leverage.
Debts don’t add neatly. Some cancel out; others are unrelated. In a complete, autarkic economy the total debt is always exactly zero: every debt is both a liability owed by someone and an equal-and-opposite asset owed to someone. But they don’t look that way on paper: the debtor must acknowledge the full amount of debt, whereas the debt-holder must acknowledge the risk of default. So on paper the total assets are less than the total liabilities. For any set of financial obligations – even if they’re not explicitly in the form of debt – this difference is what I call aggregate leverage. Aggregate leverage is one way of viewing a key difference between debt and equity: the obligation to shareholders is always exactly whatever is left on the liability side of the balance sheet and therefore changes freely in rough accord with the value of the stock to the shareholders, whereas bonds contribute to aggregate leverage. Entire segments of the economy can be financially insolvent despite possessing real assets in excess of their external financial obligations, if their aggregate leverage exceeds the difference. Even short of actual financial insolvency, excessive aggregate leverage can cause financial distress by limiting the ability to raise capital. Obviously, this is to be avoided. Regulation must ensure that it is.

3.Limit pricing feedback.
In an ideal market, prices are determined by supply and demand: in the event of a shortage or glut, the actual price is shifted back toward the price at which the market clears. In many real markets, however, this mechanism does not function adequately: the market value of the asset is whatever speculators say it is. Rather than linking producer and consumer, the market merely finds a consensus between speculators who buy at any given moment and those who sell. Such markets are subject to free-rider effects, where some participants buy or sell based not on their evaluation of the fundamental value of the asset but on what price others have been buying and selling for. Such an echo chamber amplifies noise as well as signal. If an increase (or decrease) in the price of one transaction leads to an equal total increase (or decrease) in the prices of subsequent transactions, any change will lead to a bubble or crash. Even at lower levels of feedback gain, there will be excessive volatility and persistent mis-pricing. Regulators must evaluate how financial markets function, and intervene as needed to limit pricing feedback.

4.Impose accounting requirements.
Accounting must produce comprehensible book values. These values will never reflect economic reality perfectly, and there’s an inevitable trade-off between the sophistication with which accounting mirrors economic reality and the ease with which it can be understood. However, there are also incentives to obfuscate. Voluntary accounting standards have failed to keep asset valuation comprehensible. Regulators must intervene as necessary to ensure that accounting is transparent.

5.Systematically recognize limitations of markets.
Academic economists have described numerous ways in which markets can deviate from the textbook ideal, and theorized at length about the effects of such imperfections. They also have theorized about what cannot be expected even from perfect markets. Regulators must systematically take such understanding into account when evaluating what regulation is required.

6.Regulate for systemic risk both ways.
There has been some debate as to whether it is necessary to have a new regulatory entity specifically charged with limiting systemic risk, or whether to require various regulators to assess the contribution to systemic risk from factors within the scope of their authority. The prudent course is to do both, and to include evaluation of other agencies among the responsibilities of the systemic-risk regulator.

7.Separate firms where warranted.
One of the factors implicated in the current crisis is the Gramm-Leach-Bliley Act, which repealed the mandatory separation between banks and other financial institutions provided by the Glass-Steagall Act. Such separation should be reinstated, but not necessarily in exactly the original form. Factors to consider include transparency and the incentives the combined firm and each of its components would face.

8.Provide guarantees as needed to break any contagious risk cycle.
Trouble at one firm can cause trouble at others. This can happen by outright failure, default during bankruptcy reorganization, financial leverage as described above, or effects in the real economy such as when a supplier who loses one customer no longer has the economies of scale to continue providing a product to another customer at an acceptable price. Regulators must be alert to all such possibilities and break the cycle before it becomes catastrophic, usually by providing some firms with guarantees against the losses incurred when another firm encounters trouble.

9.No implicit guarantees.
Part of the reason AIG caused as much trouble as it did is that firms counted on the government to bail it or them out if it couldn’t fulfill its obligations – and thereby created a situation where the government faced intense pressure to do so. The government must be clear about what guarantees it is providing, and what guarantees it is not. It must also ensure that it is able to follow through on its non-guarantees without catastrophic effects.

10.No promises you can’t keep.
Most of the reason AIG caused so much trouble is that it made commitments it had no means to fulfill. Firms must be forbidden from doing so.

11.No gaps in regulation.
There must be no significant players that go unregulated simply because they don’t fit any regulatory category. That may mean that some activities are entirely forbidden because they were not anticipated when regulation was devised. It may mean that regulatory schemes include catch-all categories for anything that doesn’t fit any other category. It may result in some loss of innovation, but the value of innovation is finite.

12.Policies must be predictable.
Regulatory systems will have to continue to adapt to changing circumstances and understanding. But such adaptation must occur at a limited pace so that firms will have a predictable regulatory environment in which to do business, and so that the system will not be excessively vulnerable to a wave of deregulation.

13.Too big to fail is too big to exist.
We must never be faced with another choice between a failure like that of Bear Stearns and a bailout like that of AIG. The first line of defense against this possibility should be the simplest: break up any firm that is too big to fail.

14.Risk must not just flow uphill.
Every effort at risk management has some assumptions. Those create an “unless” clause: the risk is contained unless something in the assumptions fails to hold up. When risk management is apparently perfect, nothing can go wrong – unless everything goes wrong. The classic example is the system of alliances before World War I, which seemed to have achieved “peace in our time” but in fact led to “the war to end all wars.”

15.Contagion must have a price.
Banks that receive FDIC guarantees of their deposits pay into the FDIC for the privilege of having those guarantees. The same principle must be applied in any guarantee provided under principle #8 above. The firms that receive the benefit should be required to pay the cost, to avoid distorting the incentives they face.

16.Ratings agencies must be formally part of the system.
Currently, ratings agencies can claim that they’re no different than a newspaper, exercising their rights under the First Amendment. That is not an acceptable basis for the role they play. Accounting standards should require ratings from an agency that is explicitly providing a service for a fee or otherwise taking a stake in the transaction, so that its legal position is different from that of a newspaper.

17.There has to be a reality check.
Interest rates and other rates of return theoretically reflect economic quantities such as the marginal product of capital and consumers’ marginal rate of substitution between current and future goods. In a goods or factors market, the reality check is provided by market clearing: if there’s a glut or a shortage, something is out of balance. But in financial markets when one side of the transaction is denominated in goods that don’t exist yet, this automatic reality check is not available. Regulators must devise some alternative way of providing one.

18.Profit for Wall Street is cost to Main Street.
When auto workers are making high wages, that’s widely understood to be good and bad: good because the workers get to have a high standard of living, and bad because means higher prices for consumers and lower returns for shareholders and bondholders. When the financial industry builds an Emerald City of skyscrapers, private jets, and luxurious offices, it isn’t as universally recognized that this means lower wages for workers and lower returns for Main Street investors. But it does. Maximizing “economic profit” isn’t the same as maximizing “accounting profit.”

19.If something sounds too good to be true, it probably is.
AIG was running what looked like an automatic sure-thing money-making machine. No such thing exists. Such impossible successes should be a red flag to regulators – not an automatic cause for action, because they can be mistaken about what’s too good to be true. But more often than not, further investigation will reveal that it was a wake-up call. When regulators find such a case, they must respond to it systematically, not just by addressing the particular case that drew their attention to it.