Regulators Want Banks to Rescue Themselves Next Time

Can a too-big-to-fail asset manager or insurance company hold too-big-to-fail bank debt? It's TBTFs all the way down.
Mark Carney is pleased.

I suppose we should talk about the Financial Stability Board's new global anti-too-big-to-fail proposal? Here it is. The basic theory makes sense. You have a bank. A bank is a collection of probabilities. You can estimate how much its assets are worth. But the assets may be worth much more, or much less, than you think. 1

People have claims on the bank's assets. If the assets are worth less than you think, then those claims are worth less than you think. Some of those claims are Important: It would be Bad, for the world, for some definition of the world, if those claims were worth less than you think. Bank deposits are, classically, Important; so are other bank liabilities that are necessary for the functioning of markets and the monetary system. There are various ways to guarantee against those claims being impaired when the bank loses money. One is, if the Important claims are impaired, the government will just step in and make the Important claimants whole. This is called a "bailout," more or less, and is widely perceived as Bad.

A possibly more sensible way to guarantee against impairment of Important claims is to make sure that

  1. there are a lot of Unimportant claims, 2 and
  2. the Unimportant claims are impaired before the Important ones.

This is basically how capital regulation works. Some claims on a bank's assets are called "capital," and the defining feature of capital is that if it loses value no one will be that fussed about it. 3 The people who own bank shares know what they're getting into: They know that those shares can lose value, so they're prepared for the worst.

QuickTake Too Big to Fail

One theory of bank regulation is to just make banks have a whole lot of capital and call it a day. This is the position most closely associated with Anat Admati. It is not particularly the theory of the Financial Stability Board, which has rather irritated Admati. Instead, the FSB has decided to extend the notion of capital to what it calls "Total Loss Absorbency Capacity." Capital is a subset of TLAC: Capital is the least important claimant on a bank's assets; TLAC is more of an effort to define and standardize all the Unimportant claims on the bank. Basically: capital, plus long-term debt that is subordinated to Important claims (insured deposits, repurchase agreements, derivatives liabilities, tax liabilities and any other liabilities that "cannot be effectively written down or converted into equity" if the bank runs into trouble 4 ).

Under the FSB's proposal, global systemically important banks not only can count that sort of long-term debt as TLAC; they have to have at least some of that long-term debt in their TLAC. The FSB proposal calls for TLAC of at least 16 to 25 percent of a big bank's risk-weighted assets, and at least 6 percent of its total assets. But:

To help ensure that there are sufficient resources available in resolution, there is an expectation that TLAC in the form of debt capital instruments and other TLAC-eligible liabilities that are not regulatory capital will constitute an amount equal to or greater than 33% of the Minimum Pillar 1 TLAC requirement.

So you can't just have a lot of equity capital. You need long-term unsecured debt too.

Why do it this way? One answer is what that paragraph says: You want to "ensure that there are sufficient resources available in resolution." The idea is that, if banks run out of capital entirely -- if they're technically "insolvent" -- then you can make them solvent again by poofing TLAC debt into equity. If they have no TLAC debt, only equity, then when they become insolvent you can't do that. This strikes me as an extreme bit of formalism -- why not make the requirement all equity, and seize the bank when it loses two-thirds of its equity? -- but there you have it.

Second, there's the notion that holders of long-term debt have some more ability and desire to check up on management than do holders of equity. A share of bank stock is a basically at-the-money option on the bank's assets. Option holders like volatility. A shareholder should want her bank to take risks, and should reward managers for doing so. A bondholder with a second-loss position, on the other hand, will be less inclined to take risks, and might yell at management if it takes too many risks. 5

Beyond that, though, these rules promote sort of an appealing tidiness. The point is not just to have well-capitalized banks. It's to know what would happen if a bank runs into trouble. It's to clearly define which claims are Important and which are Unimportant, and in what order. If you own TLAC debt, you know what you're getting into. You know that you're less important than derivatives counterparties. You're explicitly subordinated, and you've been explicitly promised that you won't get a bailout. The hierarchy -- common stock least important, then preferred stock and other capital instruments, then TLAC debt, and then, on the other side of the Importance divide, uninsured deposits and derivatives and ultimately insured deposits -- is becoming more explicit and standardized and fully endorsed by regulators and resolution authorities. Wherever you are in that hierarchy, you know what to expect.

The last set of rules that were supposed to solve too-big-to-fail -- last month's International Swaps and Derivatives Association initiative to add an automatic stay for derivatives contracts in bankruptcy -- had sort of a similar effect. Those derivatives claims were Important, but they were Important in an untidy way, treated differently from other Important claims that were protected by preserving the bank as a going concern. Harmonizing them with how other claims work gives a bit more predictability in how a bank failure would work.

This is not just about tidy-mindedness, though it's probably about that a little bit. Predictable risks are easier to price. If you can figure out where you are in the importance hierarchy of bank claimants, you can figure out how much you should charge banks for your money. If creditors expect to take the risk of bank failure, and charge for it, then any implication of a too-big-to-fail subsidy goes away. The costs of failure are born by your bondholders, and you pay them for doing that. 6

But there's also some social engineering required by TLAC. Instruments, by themselves, can't really be Important or Unimportant. There is no abstract answer to the question "Do we care if this instrument loses value?" Instruments are held by people or institutions, and if we care about those people or institutions, and they can't afford to lose money, then we care if the instruments lose value. 7 The only way to keep the tidy system tidy is to make sure that Unimportant bank claims don't fall into the hands of Important or sympathetic or systemic claimants. As the FSB proposal puts it:

Similarly, authorities must be confident that the holders of these instruments are able to absorb losses in a time of stress in the financial markets without spreading contagion and without necessitating the allocation of loss to liabilities where that would cause disruption to critical functions or significant financial instability.

So the FSB wants regulators to "strongly disincentivise internationally active banks from holding TLAC issued by G-SIBs." Which makes sense: If your Unimportant claims are held by global systemically important banks -- if impairing those claims would cause further bank failures -- then you can't impair them. Then they become Important again.

But then, who would hold this debt? 8 If you look at who actually holds unsecured holding company debt of banks, a lot of it is, you know, asset managers and insurance companies. There is a lot of regulatory talk about declaring some big bond fund managers "systemically important." There is regulatory action on declaring big insurance companies systemically important, much to their dismay.

It's interesting to think about how those designations would interact with the TLAC proposals. Is bank debt too risky for big insurers and asset managers to hold? Are insurers and asset managers too important to hold bank debt? If so, that would be a little weird. Someone has to ultimately bear this risk of bank failures. You need a sink, somewhere, for financial-system risk. That's an important job, and whoever does it is going to be important. But you can't protect them from the consequences of the job. Bearing those consequences is the job.

This column does not necessarily reflect the opinion of Bloomberg View's editorial board or Bloomberg LP, its owners and investors.
  1. There are some reasons to suspect upward biases. Also, here is a cool paper about financial reporting using probabilities rather than specific point numbers.

  2. There is sort of a technical definition of a lot, which is: have enough Unimportant claims that if banks lost more money than they've ever lost before they still wouldn't eat through all the Unimportant claims. That's a vague underlying theory behind capital regulation. On the other hand you want a bit of cushion:

    Some in the industry see the proposals as overkill. The new requirements will amount to as much as 25% of a bank’s risk-weighted assets, or a minimum of 6% of its total assets. The U.K.’s worst bank failure, Royal Bank of Scotland, needed government money equivalent to just 2.25% of its total assets.

  3. Another important feature is that people who own bank capital securities can't demand their money back from the bank.

  4. See section 12, "Excluded liabilities," of the TLAC termsheet (page 16 here).

  5. Anat Admati disputes these benefits, for reasons that I can't say I fully understand. (See pages 27-31 here.) She's probably right that bank creditors tend to be more dispersed, and have less power to discipline management with covenants, than say banks have over industrial companies. On the other hand her concerns seem to be aimed at "money-like" debt rather than the sort of unsecured holding-company debt that would count for TLAC, and perhaps caution prudence.

  6. From the FSB proposal:

    In terms of broader economic implications, the added funding costs associated with a TLAC requirement will lead to a reduction of the implicit public subsidy for G-SIBs. To the extent that a TLAC requirement reduces the contingent sovereign liabilities associated with bailouts, sovereign funding costs should decline as well.
  7. You can think of the Fannie Mae and Freddie Mac lawsuits in those terms (though you don't have to). Fannie and Freddie, by any reasonable definition, failed. Should the common and preferred stock go to zero? One way to answer that is to say, "Ooh it's evil hedge funds who hold that stock, it should go to zero." Another is to say, "Actually some sympathetic pensions and retirees and Ralph Nader hold it, it should be rescued." These are not legal or financial arguments but they seem to have some purchase.

  8. Or write credit default swaps against it? One thing that creeps me out, generally, is CDS on contingent capital bonds, which seems like sort of a sneaky way to reallocate claims. "You should only buy bank capital instruments if you can afford to lose your money," is kind of the regulatory thinking, so your decision to insure yourself against loss seems a bit suspect.

To contact the author on this story:
Matthew S Levine at

To contact the editor on this story:
Zara Kessler at

Before it's here, it's on the Bloomberg Terminal.