The Leverage Ratio: Neither Simple nor Sensible

The U.S. Treasury Department’s recent call for review of the calibration and design of the supplementary leverage ratio has generated renewed interest in the long-running debate over the value and appropriate role of the leverage ratio in bank capital regulation.   In light of that attention, and given the wide range of specious arguments that have been offered in support of a stringent and binding leverage ratio, we focus here on identifying key problems with the leverage ratio, and key problems in the arguments of those that support it.

First, a brief primer on the supplementary leverage ratio (SLR) itself.  There are a variety of ways to measure and calibrate the amount of capital that banks should maintain as a buffer against potential losses.  Most of these are sensible risk-based measures – that is, the amount of capital required is tethered to the actual risk of the bank’s assets and activities.  This includes the Basel Committee’s risk-based capital framework, which assigns “risk weights” to different assets (using either standardized buckets or more granular risk models, depending on the bank) that determine how much capital must be maintained against each.  It also includes the Federal Reserve’s stress testing regime, which effectively requires banks to maintain capital against the losses they might suffer under a severe economic downturn.  This risk-based approach is eminently sensible; since the point of capital is to absorb potential losses, a bank’s relative risk of loss is precisely the right point of reference in designing capital requirements.

The leverage ratio is different – it treats the risk of every asset and activity as exactly the same, regardless of its actual risk profile.  The result is an awfully poor form of prudential regulation – consider, for example, two banks of identical size, with one investing all of its funds in cash deposits at the Federal Reserve, and the other investing all of its funds in defaulted sub-prime mortgage loans.  Under a risk-based capital approach, the latter bank must hold significantly more capital – as it should.  Under a leverage ratio approach, the capital requirement for these two banks would be identical.

As we have argued in detail elsewhere (see here, here, and here), this blindness to risk is what makes the leverage ratio a particularly poor way to regulate bank capital – it is akin to setting the same speed limit for every road in the world, whether it’s a highway or a school zone.  That fatal flaw is likely why the U.S. Treasury Department recently made two recommendations regarding the leverage ratio:  (i) deducting cash, U.S. Treasury securities and customer initial margin from the leverage ratio, and (ii) reviewing and potentially revising the stringent calibration of the “enhanced” supplementary leverage ratio applicable to eight U.S. G-SIBs.  A number of commenters have expressed strident opposition to these recommendations, and in so doing have offered a range of well-worn arguments.  Because each of these arguments is flawed, we discuss and rebut each of them below.

    1.     Removing cash on deposit at central banks, U.S. Treasury securities and initial margin from the supplementary leverage ratio would undermine the primary value of the leverage ratio, which is to provide a simple and transparent non-risk-weighted capital ratio. [1]

There are three fundamental problems with this line of argument:  (i) the existing leverage ratio is neither simple nor transparent, (ii) the existing leverage ratio is not a “non-risk-weighted” measure, and (iii) deducting these riskless assets would improve, not undermine, whatever value the leverage ratio might have.

The existing SLR is already complex.  Defenders of the SLR often invoke “simplicity” as one of its cardinal virtues.  Of course, simplicity is not an unmitigated good – a simple idea can also be a dumb idea.  But even if simplicity were the ultimate goal, the simplicity of the SLR is often overstated.  It is not, as often described, a simple and pure balance sheet measure – there are plenty of devils in its details.  For example, the SLR contains pages and pages of special rules and formulas for converting derivatives, securities financing transactions and lending commitments into asset equivalents.  Because a picture is sometimes a thousand words, consider section 217.10(c)(4)(f)(2) of the Federal Reserve’s version of the SLR rules, and ask yourself how “simple and transparent” the SLR really is:

image of list

The existing SLR is not “non-risk-weighted.”  Contrary to the common assertion, the SLR is a risk-based measure – just a very poor one.  As a general matter, it assumes that all on-balance-sheet assets have exactly the same risk.  And for off-balance-sheet exposures, the complex rules described above effectively assign a wide range of different risk weights depending on the precise type of exposure at issue.  (If you have a few hours to set aside to sorting through them, you might consider reviewing Davis Polk’s very helpful 43-page summary.)  Each represents a supervisory determination as to risk.

Deducting cash, U.S. Treasury securities and initial margin for centrally cleared derivatives would improve, not undermine, the SLR.  For the reasons noted above, it’s hard to understand how deducting these effectively riskless assets makes the resulting ratio any more complex or any more opaque.  This is basic subtraction, and the result to supervisors and markets alike would be a clearer, not muddier, picture of a bank’s sensitivity to losses.  (Subtracting cash is certainly more straightforward and clearer than the complex formula we’ve highlighted above.)

Some have posited that the real risk here is the difficulty in drawing lines under the leverage ratio, but as we have previously described, this is not a legitimate concern:

It is sometimes said that deducting these assets would begin a “slippery slope.” This worry is difficult to understand – bank regulation is replete with line drawing. For example, the liquidity coverage ratio gives 100 percent credit for a central bank reserve or U.S. Treasury security as a liquid asset; this has not created a “slippery slope” whereby loans have been given 100 percent credit as a liquid asset. The Bank of England, on July 25, 2016, began deducting central bank reserves from the leverage ratio denominator for U.K. banks – and no “slippery slope” has emerged whereby it has felt the need to do so for, say, subprime loans or Greek sovereign debt.

The line drawing implicit in the U.S. Treasury’s recommendation is neither slippery nor subject to doubt.  Cash deposits at the Federal Reserve and U.S. Treasury securities pose no meaningful default risk (and if they do, their treatment under the leverage ratio will be the least of the financial system’s problems).  And the initial margin in question represents funds held on behalf of, and as agent for, customers for whom the bank has executed a centrally cleared derivative; those funds (and their risk) are ultimately the customer’s, not the bank’s.

    2.     Leverage capital had greater credibility with financial markets than risk-based capital, and is a better predictor of financial strength.[2]

As we have shown in a recent research note, once one controls for a bank’s risk-weighted capital position, the leverage ratio is actually a predictor of financial weakness, not strength. Going into the most recent financial crisis, banks with either low risk-based capital or low leverage ratios were more likely to fail – of these, the tier 1 risk-based capital ratio had the strongest ability to predict bank failure.  But if one controls for a bank’s risk-based capital ratio,  that being equal, a bank with low leverage ratio was less likely to fail.  The reason this is true is quite intuitive – if one holds equity and risk-weighted assets constant, the lower a bank’s leverage ratio, the more it must be holding low-risk assets.  In fact, of the more than 400 banks that that failed during the financial crisis, more than a quarter had leverage ratios equal to or above 10 percent.

    3.     The supplementary leverage ratio is important because relying on risk-weighting has proven entirely unsuccessful, as it is impossible to predict the future. [3]

The germ of truth in the leverage ratio is that a bank should only be able to leverage itself just so much on a risk-weighted basis, because the risk weights could be wrong for certain asset classes.  This is particularly true of assets that receive low or zero risk weights under the risk-based capital framework, where an error would have the most significant consequences; a good example is mortgage-related assets in the financial crisis.

Of course, the smart answer to this problem is to stress test those assets, and ensure that banks maintain capital requisite to cover their likely losses in a severe downturn, even when recent history has shown low loss experience and thereby dictates low risk weights.  Banks perform stress tests on a regular basis, and the CCAR and DFAST stress tests are a central component of post-crisis reform.  In this sense, the CCAR and DFAST process is as much a stress test of the risk-based capital framework and its potential weaknesses as it is of the banks’ capital adequacy.

The leverage ratio is a much clumsier answer.  Again, it assumes that every asset has the same credit risk:  that the risk of default on a subprime mortgage is the same as the risk of default on a U.S. Treasury security – or even the risk of default on cash in your wallet (e.g., the risk that you will present cash at the grocery store and be told you’ll only be given partial credit for it).  This is, of course, an assumption that is deeply counterfactual, to the point of being nonsensical.  And any banker who assessed the risk of loss as the same on every loan would and should be fired.

A leverage ratio can serve as low cost insurance against a breakdown in modeled risk weights if it is calibrated as a backstop measure.  Thus, an effective leverage ratio ought to be calibrated such that it binds only in the relatively rare circumstance where a bank has disproportionately large holdings of assets that, while receiving low or zero risk weights under the risk-based approach, might perform poorly under stress.  We note that while such use cases may from time to time exist, they do not include one where a bank holds large amounts of cash deposits at the Federal Reserve or U.S. Treasury securities, which one should safely assume will be the last asset classes standing under any kind of stress.

Unfortunately, U.S. regulators have calibrated the SLR at a very high level, especially for U.S. G-SIBs, even as stress testing has reduced its marginal benefits and usefulness.  Worse yet, the leverage ratio comes with two significant costs, if calibrated too high.  First, it increases the cost of banking services, particularly capital markets and custody services.  Second, it penalizes the holding of low-risk assets and rewards (on a relative basis) the holding of high risk assets, thereby creating systemic risk.

So, think of the leverage ratio as a particularly unstable form of dynamite – a tool that can be useful in some contexts, but one that needs to be handled very carefully.

[1] Examples of this line of argument include recent articles by Nellie Liang and Gregg Gelzinis.

[2] A recent example of this line of argument is offered by FDIC Chairman Gruenberg in his June 22, 2017 testimony before the Senate Banking Committee.

[3] For an example of this line of argument, see the statements of FDIC Vice Chairman Hoenig.