VoxEU Column Financial Markets

Market-based bank capital regulation

Today’s regulatory rules – especially the ineffective capital requirements – have led to costly bank failures. This column proposes a new, robust approach that uses market information but does not depend upon markets being ‘right’. Under the proposed regulatory system (i) bank losses are credibly borne by the private sector (ii) systemically important institutions cannot collapse suddenly; (iii) bank investment is counter-cyclical; and (iv) regulatory actions depend upon market signals. One key innovation is ‘Equity Recourse Notes’ that gradually ‘bail in’ equity as needed. These are superficially similar to, but fundamentally different from, 'CoCos'.

Current bank regulatory capital systems fail in three important ways:

  • Regulatory capital measures bear little relation to firms’ financial health.

Financial institutions ranging from Lehman, Wachovia, and Citigroup in the US to Bankia and Dexia in Europe were rated as well capitalized on the day they failed, got bailed out, or were acquired. Just two months before it requested a €90 billion bailout guarantee, Dexia passed the 2011 European stress tests with flying colours (EBA 2011). Its core equity projected to fall no lower than €15.3 billion (about 10%) in the ‘adverse scenario’. Regulatory rules distort both investment decisions and risk-management.

  • ‘Too big to fail’ means the ‘property rights’ in bank losses are not clearly and credibly allocated.

We don’t know which debt holders will be asked to contribute how much to rescue insolvent institutions. While efforts are being made to make bankruptcy less costly and so more plausible (through e.g. living wills and bail-in), it is hard to know whether politicians will allow these mechanisms to operate as planned. It will always be more expedient to push off a default, and sometimes it might even be the right policy. But the consequence is that tail-risk gets shifted, both explicitly and implicitly, to citizens who are ill-equipped to bear it. And because regulatory capital rules allow banks to overstate assets, and encourage them to gamble on ‘mean reversion’ in asset prices, taxpayers’ losses can be enormous.

  • The current system is pro-cyclical.

Distressed banks (those with little market capital relative to their regulatory capital and liabilities) find it unappealing or impossible to raise the new equity required to make new risky loans, because so much of any equity raised will simply go to reducing the expected losses of creditors, including the government who is insuring the deposits. That is, the current regulatory system creates a severe ‘debt overhang’ problem that in bad times acts just like a tax on new investment.1  As a result, troubled banks contract, as in Europe 2008-present.

Easy solutions won’t work

Simply doubling or trebling capital requirements won’t do. For example, in 2008-11 the US Federal Deposit Insurance Corporation (FDIC) lost money on 413 bank failures. Say that those banks – which required 6% core tier 1 regulatory equity to be classified as ‘well capitalized’ – each held an extra 14% of assets in cash, but no extra debt on the day they failed. This infusion would have been insufficient to cover losses in 372 (90%) of these cases.2  Of course most of these banks were relatively simple; more complex large banks might have better risk management, but also more scope for trouble.

Furthermore, high capital requirements have costs too, creating incentives to move assets to ‘shadow banks’ and other differently regulated institutions. Moreover, merely changing capital requirements would do nothing about the problems of pro-cyclicality, or the pressures on regulators to relax requirements in a recession. We need to transition to a fundamentally more robust system while banks are temporarily less reliant on government support.

Our solution

Our solution is based on two rules.

  • First, any systemically important financial institution (SIFI) that cannot be quickly wound down must limit the recourse of non-guaranteed creditors to assets posted as collateral plus equity plus unsecured debt that can itself be converted into equity -- so these creditors have some recourse but cannot force the institution into re-organization.
  • Second, any debt guaranteed by the government, such as deposit accounts, must, in the long run, be backed by government-guaranteed securities.

The second leg of our reforms is a pipe-dream for now. So our interim objective is to ‘ring fence’ government-guaranteed deposits separate from all other liabilities in a manner broadly along the lines of current policy proposals such as those of the U.K. Independent Commission on Banking (2011).3   We would however require their collateralization by assets that are haircut by percentages similar to those applied by lenders (including central banks and commercial banks themselves!) to secured borrowers.4

Part one: Replace all unsecured debt with ERNs

Under our plan we would first have banks replace all (non-deposit) existing unsecured debt with ‘Equity Recourse Notes’ (ERNs).

ERNs are superficially similar to contingent convertible debt (‘CoCos’) but are fundamentally different.

ERNs would be long-term bonds with the feature that any interest or principal payable on a date when the stock price is lower than a pre-specified price would be paid in stock at that pre-specified price. The pre-specified price would be required to be no less than (say) 25% of the share price on the date the bond was issued For example, if the stock were selling at $100 per share on the day a bond was issued, and then fell below $25 by the time a payment of $1000 was due, the firm would be required to pay the creditor 1000 ÷ 25 = 40 shares of stock in lieu of the payment. If the stock rebounded in price, future payments could again be in cash.

Crucially, for ERNs, unlike CoCos:

i. Any payments in shares are at a pre-set share price, so ERNs are stabilizing because that price will always be at a premium to the market;5

  • That is, ERNs have the opposite effect to so-called "death-spiral" bonds that convert at a discount to the current price. Because ERN conversions are always for a fixed number of shares at prices above the then current share price, every conversion transfers wealth to current shareholders, and so shores up the share-price. In effect, banks buy puts from lenders every time they sell an ERN, which transfers risk from shareholders to ERN holders, and so reduces share price volatility.6

ii. Conversion is triggered by market prices, not regulatory values—removing incentives to manipulate regulatory measures, and making it harder for regulators to relax requirements;
iii. Conversion is payment-at-a-time, not the entire bond at once (because ERNs become equity in the states that matter to taxpayers, they are, for regulatory purposes, like equity from their date of issuance, so there is no reason for faster conversion) -- further reducing pressures for ‘regulatory forbearance’ and also largely solving a ‘multiple equilibria’ problem raised in the academic literature;
iv. We would replace all existing unsecured debt with ERNs, not merely a fraction of it.

These distinctions eliminate the flaws of standard contingent convertible debt.7Importantly, this design also ensures, as we explain below, that ERNs become cheaper to issue when the stock price falls, creating counter-cyclical investment incentives when they are most needed.

Our plan would also require a subtle but crucial transformation of secured (i.e., collateralized) borrowing by banks. Currently a secured creditor has a claim against the posted assets plus an unsecured claim against the bank for any shortfall, if things go wrong. If there is a good chance that the government will bail out a failing bank, the terms of the loan will be based partially on the government’s credit rather than on the quality of the bank’s. The solution is to limit the recourse that a creditor would have, beyond the posted collateral, to either shares of stock or ERNs.

Part two: Deposit accounts as money-market funds

In our ideal world, deposit accounts would follow a money market fund model. Government guaranteed accounts would be like existing money market funds that invest in government-guaranteed debts, along the lines of 100% reserve proposals.

Currently, depositors in effect receive short-term government-guaranteed debt, acquired from banks that obtain it in return for unsecured bank debt plus mispriced, cheap, deposit insurance. We would eliminate the ‘middle man’, so that depositors directly hold loans from the government.8

Creditors could still acquire short-term unsecured (i.e., non-collateralized) bank debt in much the way they can acquire short-term debt from, say, mortgage securities. Investment trusts could purchase ERNs and pool and tranche them, issuing more senior and shorter-term claims to those who want them, The difference is that in a panic, which might cause an investment vehicle to sell some of its bonds to pay short-term claims, losses would be borne by those who took levered, junior claims in the trust without any short-term repercussions for the underlying banks’ financial condition. Furthermore, investors who wanted to reduce or eliminate the tail risk of their ERNs could do so by buying equity puts. They could transfer the risk to any willing buyer – just not to the taxpayer.

Some of the senior unsecured claims generated by such trusts might be held in non-guaranteed money market accounts, operated under rules akin to those proposed by the US S.E.C. in 2012, with floating net asset values so that it would quickly become clear to investors that the accounts had some risk like in a very short-term bond fund.

Similar principles could be applied to the funding of derivatives and other potential liabilities.

Generating countercyclical pressures

Our system generates strong countercyclical pressures:

  • Debt payments automatically convert to equity in times of stress, so automatically repair the capital structure
  • ERNs become cheaper to issue when the stock price falls. If, e.g., the stock price declines from 100 to 40, new ERNs can be issued with a conversion price of 10 instead of 25 – so the new bonds will only suffer losses after the old bonds have already taken a 60% haircut. The more the stock price declines, the more senior new issues can be; if a stock hits a low new, ERNs will be senior to all other unsecured capital and so especially cheap to issue
  • Issuing new senior debt (ERNs) can send a better signal about the company’s prospects than selling assets, because issuing ERNs is relatively cheap; by contrast, raising new funds in the existing system is a worse signal than selling off assets, because of the need to raise equity to maintain its regulatory-capital ratio
  • The existing system’s ‘debt-overhang’ problem that acts like an investment tax in bad times is reversed.

Because all creditors’ final recourse is to ERNs – which become equity in bad world-states – our system alleviates liquidity as well as solvency problems. So our system also mitigates ‘downward spirals’ and liquidity crises, while allowing poorly-run firms to gradually decline and fail or recover.

The plan thus respects an important ‘time-consistency’ constraint – regulators and politicians cannot be counted on to permit sudden failures and, also importantly, institutions know this: our solution allows banks and systemically important institutions to fail, but with a whimper rather than a bang.


Many proposals add ever-more elaborate regulations to an already baroque regulatory system – one that is already unmanageable. We propose instead to make things much simpler.

ERNs are a counterweight to pro-cyclicality. They make capital raising – and therefore lending – easier rather than harder in recessions. Counter-cyclicality also increases the credibility of the plan, because there will be no incentive to scrap it in bad times. Jettisoning complex capital rules, and simply transferring tail risk back where it belongs (with private investors) takes taxpayers off the hook and ensures that banks with profitable opportunities can use them. ERNs also reduce share-price volatility relative to conventional debt.

In short, our system eliminates distortionary incentives for regulatory arbitrage and forbearance, facilitates counter-cyclical raising of unsecured capital, and clearly and credibly assigns losses to private investors where they belong.

While we rely on markets to determine banks’ risk capital requirements, our system is robust to the market being wrong, or less accurate on average than regulators’ or banks’ internal models. By contrast, current regulatory models will often lose money, and perhaps cause a crisis, if markets are right.

Full details of our plan are in Bulow and Klemperer (2013). It need not (and will not!) be implemented all at once. In the short run, some of the features of our ERNs could improve the existing design of CoCos; in the medium-term, a full transition to ERNs, ideally in conjunction with ‘ring-fencing’, would substantially stabilise the financial system.


Bulow, J, and P Klemperer (2013), Market-Based Bank Capital Regulation, mimeograph.

EBA (2011). ‘European banking authority 2011 EU-wide stress test aggregate report’, European Banking Authority, http://www.eba.europa.eu/documents/10180/15935/EBA_ST_2011_Summary_Report_v6.pdf/54a9ec8e-3a44-449f-9a5f-e820cc2c2f0a

Independent Commission on Banking, ‘Final Report: Recommendations‘, September 2011.

1 By ‘debt overhang’ problem, we mean a requirement that new investments be funded with securities that are on average sufficiently junior that wealth is transferred from existing shareholders to creditors (by transferring expected costs of default in this way). In particular, regulatory capital requirements may force a bank wishing to make a new investment to increase its share capital by a sufficiently greater fraction than the fraction by which it increases its debt that the total value of its deposit insurance falls, even though its guaranteed outstanding debt increases—see Bulow and Klemperer (2013).

2 Source: FDIC Historical Statistics on Banking, as of December 10, 2012.

3 In our system, ‘ring fencing’ is not so much a way of limiting the activities of commercial banks as a way of reducing banks’ reliance on deposit insurance, and ensuring that newly-issued ERNs (as described below) will be senior securities in bad times.

4 See for example the Federal Reserve Discount Window and Payment System Risk Collateral Margins Table and the Bank of England Summary of haircuts eligible for the Bank’s lending operations.

5  That is, the number of shares that ERN holders receive in any conversion is computed using a share price above the price at the time of conversion (the opposite of “death-spiral” bonds).

6 ERNs also automatically provide additional protection against downward share-price spirals: in particular, we explain below that ERNs become cheaper to issue when the stock price falls, so raising new money at low stock prices is both easy and also transfers wealth to shareholders, further bolstering the stock price; moreover, our full proposal means companies never have to pay cash so cannot be threatened by liquidity crises. More details are given in Bulow and Klemperer (2013).

7  In particular this design prevents downward share-price “death” spirals—see (i) above.

8 Our interim objective of tight ring-fencing would not achieve this, but would theoretically require banks to secure deposits with a pool of assets against which other lenders would be willing to provide enough cash to fully back deposits. There would be inevitable pressure to relax regulatory haircuts to below-market levels. In the longer run we would prefer to see the government out of the secured lending business as well, except in acting as a lender of last resort.

3,045 Reads