Two Cheers for the Leverage Ratio

Bank Capital, Basel III, leverage ratio, Risk Weighted Assets, unlimited liability
Modified version of the 'Community Chest' card from the game Monopoly, by Hasboro.

BeautyContestIn a previous blog posting, I suggested that there is no case for capital adequacy regulation in an unregulated banking system.  In this ‘first-best’ environment, a bank’s capital policy would be just another aspect of its business model, comparable to its lending or reserving policies, say.  Banks’ capital adequacy standards would then be determined by competition and banks with inadequate capital would be driven out of business.

Nonetheless, it does not follow that there is no case for capital adequacy regulation in a ‘second-best’ world in which pre-existing state interventions — such as deposit insurance, the lender of last resort and Too-Big-to-Fail — create incentives for banks to take excessive risks.  By excessive risks, I refer to the risks that banks take but would not take if they had to bear the downsides of those risks themselves.

My point is that in this ‘second-best’ world there is a ‘second-best’ case for capital adequacy regulation to offset the incentives toward excessive risk-taking created by deposit insurance and so forth.  This posting examines what form such capital adequacy regulation might take.

At the heart of any system of capital adequacy regulation is a set of minimum required capital ratios, which were traditionally taken to be the ratios of core capital[1] to some measure of bank assets.

Under the international Basel capital regime, the centerpiece capital ratios involve a denominator measure known as Risk-Weighted Assets (RWAs).  The RWA approach gives each asset an arbitrary fixed weight between 0 percent and 100 percent, with OECD government debt given a weight of a zero.  The RWA measure itself is then the sum of the individual risk-weighted assets on a bank’s balance sheet.

The incentives created by the RWA approach turned Basel into a game in which the banks loaded up on low risk-weighted assets and most of the risks they took became invisible to the Basel risk measurement system.

The unreliability of the RWA measure is apparent from the following chart due to Andy Haldane:

Figure 1: Average Risk Weights and Leverage

Average RWAs

This chart shows average Basel risk weights and leverage for a sample of international banks over the period 1994–2011.  Over this period, average risk weights show a clear downward trend, falling from just over 70 percent to about 40 percent.  Over the same period, bank leverage or assets divided by capital — a simple measure of bank riskiness — moved in the opposite direction, rising from about 20 to well over 30 at the start of the crisis.  The only difference is that while the latter then reversed itself, the average risk weight continued to fall during the crisis, continuing its earlier trend.  “While the risk traffic lights were flashing bright red for leverage [as the crisis approached], for risk weights they were signaling ever-deeper green,” as Haldane put it: the risk weights were a contrarian indicator for risk, indicating that risk was falling when it was, in fact, increasing sharply.[2]  The implication is that the RWA is a highly unreliable risk measure.[3]

Long before Basel, the preferred capital ratio was core capital to total assets, with no adjustment in the denominator for any risk-weights.  The inverse of this ratio, the bank leverage measure mentioned earlier, was regarded as the best available indicator of bank riskiness: the higher the leverage, the riskier the bank.

These older metrics then went out of fashion.  Over 30 years ago, it became fashionable to base regulatory capital ratios on RWAs because of their supposedly greater ‘risk sensitivity.’  Later the risk models came along, which were believed to provide even greater risk sensitivity.  The old capital/assets ratio was now passé, dismissed as primitive because of its risk insensitivity.  However, as RWAs and risk models have themselves become discredited, this risk insensitivity is no longer the disadvantage it once seemed to be.

On the contrary.

The old capital to assets ratio is making a comeback under a new name, the leverage ratio:[4] what is old is new again.  The introduction of a minimum leverage ratio is one of the key principles of the Basel III international capital regime.  Under this regime, there is to be a minimum required leverage ratio of 3 percent to supplement the various RWA-based capital requirements that are, unfortunately, its centerpieces.

The banking lobby hate the leverage ratio because it is less easy to game than RWA-based or model-based capital rules.  They and their Basel allies then argue that we all know that the RWA measure is flawed, but we shouldn’t throw out the baby with the bathwater.  (What baby? I ask. RWA is a pretend number and it’s as simple as that.)  They then assert that the leverage ratio is also flawed and conclude that we need the RWA to offset the flaws in the leverage ratio.

The flaw they now emphasize is the following: a minimum required leverage ratio would encourage banks to load up on the riskiest assets because the leverage ratio ignores the riskiness of individual assets.  This argument is commonly made and one could give many examples.  To give just one, a Financial Times editorial — ironically entitled “In praise of bank leverage ratios” — published on July 10, 2013 stated flatly:

Leverage ratios …  encourage lenders to load up on the riskiest assets available, which offer higher returns for the same capital.

Hold on right there!  Those who make such claims should think them through: if the banks were to load up on the riskiest assets, we first need to consider who would bear those higher risks.

The FT statement is not true as a general proposition and it is false in the circumstances that matter, i.e., where what is being proposed is a high minimum leverage ratio that would internalize the consequences of bank risk-taking.  And it is false in those circumstances precisely because it would internalize such risk-taking.

Consider the following cases:

In the first, imagine a bank with an infinitesimal capital ratio.  This bank benefits from the upside of its risk-taking but does not bear the downside.  If the risks pay off, it gets the profit; but if it makes a loss, it goes bankrupt and the loss is passed to its creditors.  Because the bank does not bear the downside, it has an incentive to load up on the riskiest assets available in order to maximize its expected profit.  In this case, the FT statement is correct.

In the second case, imagine a bank with a high capital-to-assets ratio.  This bank benefits from the upside of its risk-taking but also bears the downside if it makes a loss.  Because the bank bears the downside, it no longer has an incentive to load up on the riskiest assets.  Instead, it would select a mix of low-risk and high-risk assets that reflected its own risk appetite, i.e., its preferred trade-off between risk and expected return.  In this case, the FT statement is false.

My point is that the impact of a minimum required leverage ratio on bank risk-taking depends on the leverage ratio itself, and that it is only in the case of a very low leverage ratio that banks will load up on the riskiest assets.  However, if a bank is very thinly capitalized then it shouldn’t operate at all.  In a free-banking system, such a bank would lose creditors’ confidence and be run out of business.  Even in the contemporary United States, such a bank would fall foul of the Prompt Corrective Action statutes and the relevant authorities would be required to close it down.

In short, far from encouraging excessive risk-taking as is widely believed, a high minimum leverage ratio would internalize risk-taking incentives and lead to healthy rather than excessive risk-taking.

Then there is the question of how high ‘high’ should be.  There is of course no single magic number, but there is a remarkable degree of expert consensus on the broad order of magnitude involved.  For example, in an important 2010 letter to the Financial Times drafted by Anat Admati, she and 19 other renowned experts suggested a minimum required leverage ratio of at least 15 percent — at least five times greater than under Basel III — and some advocate much higher minima.  Independently, John Allison, Martin Hutchinson, Allan Meltzer and yours truly have also advocated minimum leverage ratios of at least 15 percent.  By a curious coincidence, 15 percent is about the average leverage ratio of U.S. banks at the time the Fed was founded.

There is one further and much under-appreciated benefit from a leverage ratio.  Suppose we had a leverage ratio whose denominator was not total assets or some similar measure.  Suppose instead that its denominator was the total amount at risk: one would take each position, establish the potential maximum loss on that position, and take the denominator to be the sum of these potential losses.  A leverage-ratio capital requirement based on a total-amount-at-risk denominator would give each position a capital requirement that was proportional to its riskiness, where its riskiness would be measured by its potential maximum loss.

Now consider any two positions with the same fair value.  With a total asset denominator, they would attract the same capital requirement, independently of their riskiness.  But now suppose that one position is a conventional bank asset such as a commercial loan, where the most that could be lost is the value of the loan itself.  The other position is a long position in a Credit Default Swap (i.e., a position in which the bank sells credit insurance).  If the reference credit in the CDS should sharply deteriorate, the long position could lose much more than its current value.  Remember AIG! Therefore, the CDS position is much riskier and would attract a much greater capital requirement under a total-amount-at-risk denominator.

The really toxic positions would be revealed to be the capital-hungry monsters that they are.  Their higher capital requirements would make many of them unattractive once the banks themselves were to made to bear the risks involved.  Much of the toxicity in banks’ positions would soon disappear.

The trick here is to get the denominator right.  Instead of measuring positions by their accounting fair values as under, e.g., U.S. Generally Accepted Accounting Principles, one should measure those positions by how much they might lose.

Nonetheless, even the best-designed leverage ratio regime can only ever be a second-best reform: it is not a panacea for all the ills that afflict the banking system.  Nor is it even clear that it would be the best ‘second-best’ reform: re-establishing some form of unlimited liability might be a better choice.

However, short of free banking, under which no capital regulation would be required in the first place, a high minimum leverage ratio would be a step in the right direction.


[1] By core capital, I refer the ‘fire-resistant’ capital available to support the bank in the heat of a crisis.  Core capital would include, e.g., tangible common equity and some retained earnings and disclosed reserves.  Core capital would exclude certain ‘softer’ capital items that cannot be relied upon in a crisis.  An example of the latter would be Deferred Tax Assets (DTAs).  DTAs allow a bank to claim back tax on previously incurred losses in the event it subsequently returns to profitability, but are useless to a bank in a solvency crisis.

[2] A. G. Haldane, “Constraining discretion in bank regulation.” Paper given at the Federal Reserve Bank of Atlanta Conference on ‘Maintaining Financial Stability: Holding a Tiger by the Tail(s)’, Federal Reserve Bank of Atlanta 9 April 2013, p. 10.

[3] The unreliability of the RWA measure is confirmed by a number of other studies.  These include, e.g.: A. Demirgüç-Kunt, E. Detragiache, and O. Merrouche, “Bank Capital: Lessons from the Financial Crisis,” World Bank Policy Research Working Paper Series No. 5473 2010); A. N. Berger and C. H. S. Bouwman, “How Does Capital Affect Bank Performance during Financial Crises?” Journal of Financial Economics 109 (2013): 146–76; A. Blundell-Wignall and C. Roulet, “Business Models of Banks, Leverage and the Distance-to-Default,” OECD Journal: Financial Market Trends 2012, no. 2 (2014); T. L. Hogan, N. Meredith and X. Pan, “Evaluating Risk-Based Capital Regulation,” Mercatus Center Working Paper Series No. 13-02 (2013); and V. V. Acharya and S. Steffen, “Falling short of expectation — stress testing the Eurozone banking system,” CEPS Policy Brief No. 315, January 2014.

[4] Strictly speaking, Basel III does not give the old capital-to-assets ratio a new name.  Instead, it creates a new leverage ratio measure in which the old denominator, total assets, is replaced by a new denominator measure called the leverage exposure.  The leverage exposure is meant to take account of the off-balance-sheet positions that the total assets measure fails to include.  However, in practice, the leverage exposure is not much different from the total assets measure, and for present purposes one can ignore the difference between the two denominators.  See Basel Committee on Banking Supervision, “Basel III: A global regulatory framework for more resilient banks and banking systems.”  Basel: Bank for International Settlements, June 2011, pp. 62-63.

  • Julien Noizet

    Pretty good post.
    I wrote a post along those lines a couple of years ago:

    I agree that leverage ratios are a 'second best' and nothing more. It is still unclear what sort of capital instruments should be used (only pure equity? what about hybrid debt or preference shares that behave very similarly to equity?) or what should be included in the denominator (off-balance sheet exposures like undrawn credit lines?). Derivatives netting also remains a tricky issue.
    Your 'potential maximum loss' proposal addresses some of those issues.

    Unfortunately, the Basel 3 leverage ratio does not work this way and reintroduced the equivalent of risk-weights for off-balance sheet exposures, called CCF (credit conversion factors), which give 'weights' to various sorts of banking products, thereby reproducing the sort of behaviour that were the flaws of RWAs in the first place: banks tend to avoid high-CCF products.

    See the list of CCF pages 18 and 19 here:

    • Kevin Dowd

      Thank you Julien!
      First off, I have enjoyed your posts, which are uncommonly good, including the one you pointed me to in your opening comments.
      Regarding the capital numerator in the leverage ratio, I take the view that anything gameable is unreliable in a crisis and therefore does not belong in a true core capital measure. So, eg, even retained earnings is gameable. I would exclude any hybrids and am skeptical even of preferreds. We are then left with common tangible equity.
      Netting and hedge effectiveness I regard as huge problems, which are very poorly understood, and which greatly undermine any reported numbers. Not to mention Level 2 ('mark to model') and Level 3 ('mark to myth') valuations.
      Regarding my suggested 'potential maximum loss' exposure measure: this helps to address what is needed in the denominator, but there is much more to be said on this issue and I could only really only hint at these in this posting, given space constraints. You are absolutely right that the Basel III leverage exposure denominator is very poor and introduces the equivalent of new risk-weights, thereby taking us back to the very RWA problems that the darn leverage exposure measure was meant to take us away from in the first place! You just can't make it up :-)

  • Per Kurowski

    Two comments on a topic I have been opining the most on since way before this century.

    First: The problem with RWA is that risk weights are mindboggling stupid. Basel II assigned a risk weight of 20% to AAA rated assets and 150% to below BB- rated assets. And I just ask: Kevin what do you think is more dangerous for the stability of the banking system, AAA rated assets or below BB- rated assets? Which of these assets could cause the build up of dangerous excessive exposures?

    Second: The problem with leverage ratio is that while you do not get rid of the risk-weighted capital requirements, the increased pressure on capital will increase the distortions on the margin caused by the risk weights. The drowning pool syndrome.

    • Kevin Dowd

      Thank you, Per!
      I agree with you that the risk-weights are mindbogglingy stupid – nice way to put it too. As for your question about whether triple A or

      • Per Kurowski

        Banks consider perceived credit risk when deciding on interest rates (risk premiums) and size of exposure. So when regulators clear, again, for the same perceived risk, in the capital then perceived credit risk has got too much consideration. And any risk, even if perfectly perceived, causes the wrong actions if excessively considered.

        Also Kevin… rest assure what’s AAA rated is always more dangerous, since banks will never ever build up excessive exposures to what is below BB- rated. Of course what’s AAA can turn into below BB-, but that’s another story, and that risk should have been considered when placing an AAA rated assets on the balance sheet.

        • Kevin Dowd

          The way the you express the AAA vs BB- issue in your second paragraph I do agree with: banks would not normally build up excessive exposure to below BB- rated junk ex ante, but could be left holding a lot of it ex post as misrated AAA securities get revealed to be the junk they really are. I take the point that there is a lot of AAA rubbish out there deliberately designed to game the far-from-adequate rating system.

          • Per Kurowski

            Capital should be there to cover for ex-post unexpected losses of any nature… but the regulators made the requirements dependent on ex ante expected perceived credit losses… loony!

            Credit risk are already cleared for by banks in interest rates and size of exposure, and so to clear for that same risk again in the capital is crazy… since any risk, even if perfectly perceived, provokes the wrong actions, if excessively considered.

          • Kevin Dowd

            "Capital should be there to cover for ex-post unexpected losses of any nature …" Very nicely put.
            Regarding your second para, my view is that if they are suitably incentivised, eg, by appropriate personal liability, bankers would then look for rules of thumb to determine internal capital allocations to cover the risks of different positions. This then works more or less well, albeit not perfectly, until the government and the regulators get involved and start imposing capital allocation rules that are counterproductive in various ways. Then it all goes horribly wrong.

          • Per Kurowski

            Is it not suitably incentivised enough, to provide shareholders with the highest risk adjusted returns on equity? What more should a banker think of? The problem was that the regulators manipulated the risk adjusted ROEs in favor of The Safe, the AAArisktocracy

          • Kevin Dowd

            It is NOT enough to provide shareholders with the highest risk-adjusted returns on equity. Leaving aside the troublesome question of you do the risk-adjustment, one also has to consider the incentives that bankers face and in particular, the question of whether there are public subsidies to take excessive risks, by which I mean to take risks that other people have to bear.
            One of the advantages of a high minimum leverage ratio is that it helps to rectify this government-created 'externality'. But under the best system, free banking, bankers would operate with strong personal incentives to take responsible risks, ideally, unlimited liability. Whether they then use some in-house risk-adjusted return metric, I don't care and would be entirely up to them.

          • Per Kurowski

            No Kevin. You would want your bankers to maximize their risk adjusted returns on equity. What you do not want is for regulators to distort in such a way as to telling bankers, if only you lend to the "safe" we will allow you to leverage much more. Bankers operating with strong personal incentives would still have to maximize risk adjusted returns on equity… or what other compass do you have in mind?

          • Kevin Dowd

            No and yes, Per!
            No, as far as I am concerned, bankers can do anything they want, provided only that they operate under a suitable legal regime and have the freedom to make their own decisions and are liable for the mistakes they make. If they wish to maximise risk-adjusted return on equity – by which we really should say, if they wish to maximise expected risk-adjusted return on equity – then as far as I am concerned, they are welcome to. However, this particular decision-making rule is only one of many, is only 50+ years old, and there are many alternatives. Consider, any of the standard banking textbooks from the period before modern finance theory, which offer advice about good lending practices, for example. I would not wish any bankers to forget such sound advice.
            Yes, we do NOT want regulators sticking their oars in, imposing stupid rules that have no business sense and which only create destructive incentives that undermine the system, and often in unintended ways too.

          • Per Kurowski

            Kevin, give me any best good lending practice that does not include banks maximizing the risk adjusted return on equity. The problem, your problem, is being able to accept the possibility that the regulators in the Basel Committee had absolutely no idea of what they were doing… I know its hard.

            To help you along the road try to find when current regulators defined the purpose of our banks… you won’t find it… and then reflect on how you regulate anything without defining its purpose.


          • Kevin Dowd

            My dear Per, I already DID give you a whole family of such examples, namely everything written about bank lending (i.e., the standard textbooks) written before, say, 1950. Isn't that plenty enough?

            My point is that your decision rule is but one decision rule among many, and there is absolutely no requirement (at least on my part) that any particular banker MUST use it. It is also model-specific – how do you determine risk-adjusted? – which makes me smell a rat. Instead, bankers have to make those decisions for themselves. As far as I am concerned, let them get on with it. Good luck if they get that right, sod' em if they don't.

            But I also think you are mixing up two different issues, to wit: what banks should do vs. the legal/regulatory framework under which they operate or should operate.

            In my posting, I was trying to draw attention to the latter, and avoid trying to give business advice to bankers. This is a most important distinction. I know enough about the latter to know that I should generally not comment about it, except maybe to expose what I believe to be bad thinking about it.

            My friend, please allow me a little rant of my own: I do wish you would not use the terminology of banks maximising risk-adjusted returns because that is simply impossible. I realise that this language is standard practice, but it is fatuous nonetheless. Needless to say, I do not intend this is as a personal criticism, but as a swipe against widespread bad practice that your choice of language here represents.

            The reason why this makes no sense is simple: you cannot predict the future. Crystal ball 101: the return next period is random, risk-adjusted or not. Ex ante, this period, when I make the decision about which risks to take, I do not know what the outcome will be. If I did, then I would have a crystal ball that actually works and I would be writing to you from my Caribbean island. Unfortunately, I don't and I am not. It is impossible to maximise the realised value of a future random event. Instead, all one can ever do is maximise something like the expectation of a future random event. There are variants on this theme but lets forget those for now, but you have to get the measure theory correct. If one insists on using this risk-adjusted return etc. approach, you should always qualify the language, typically by adding the modifier "expected".

            Otherwise, one is suggesting a decision rule that is impossible to implement. What is the point of that, let alone suggesting that it is the ONLY decision rule they should use? In any case, I was trying to discuss the regulatory system …

            Let me turn to your other points:

            "The problem, your problem, is being able to accept the possibility that the regulators in the Basel Committee had absolutely no idea of what they were doing… I know its hard."

            You are way too kind to me in suggesting that I have only one problem, but let me thank you for that and move on. I defer to your experience but I think the issues are more involved than you suggest. First off, I believe they DID have no idea what they were really doing, at in terms of the longer term consequences of their actions. But, second, there are much deeper issues here: groupthink, self-delusion, they-know-best, regulatory capture, and I cannot begin to properly get into these issues in a response to a blog posting: I would have to write a book about it.

            "To help you along the road try to find when current regulators defined the purpose of our banks… you won’t find it… and then reflect on how you regulate anything without defining its purpose."

            So what is your point? Surely you are not suggesting that regulators need some clear purpose? My observation of them in action over decades clearly shows that they don't. They just regulate and make up 'purposes' and 'objectives' along the way, but even they don't take their own purposes/objectives seriously so I don't either.

          • Per Kurowski

            Dear Kevin, we are not that far away in our perceptions, but let me be clear that when I refer to bankers having to maximize risk adjusted returns on equity I always refer to "perceived" maximized risk-adjusted returns on equity, not to de facto ex post disasters.

            At the end of the day, it all comes down to this, do you believe below BB- rated assets are more dangerous to the bank system than AAA rated… because that is what the Basel Committee believed,

          • Kevin Dowd

            Dear Per,

            Thank you for this and I do believe you are right: we are not that far apart.

            And, yes, the Basel Committee clearly believed that AAA assets are safer than those rated below BB- – and we learned the hard way that that is not a safe assumption to make when the rating system is broken and the bankers are gaming it with AAA-rated securitisations that can be worse than junk.

            I for one (and I dare say you would agree) am not convinced that they (Basel) have learned this lesson. In fact, I am convinced that they have not.

            Thank you for an interesting discussion, Per :-)

          • Per Kurowski

            Kevin you write about credit ratings being wrong but let me assure you that, even perfect credit ratings, if excessively considered, lead to the wrong actions. Credit ratings were already cleared for with interest rates and size of exposures, and so to clear for it again in the capital, wrecked the whole system of bank credit allocation to the real economy

      • George Selgin

        I didn't know you were such a fan of equality, Kevin!

        • Kevin Dowd

          Yes indeed! This is the only thing you will ever see from me in favor of it too. I particularly like the artistic effect of all that """" stuff. But in fact, I can't really take credit. What I actually wrote was this – which looked fine in draft form but then got magically transformed when I pressed the 'post' button:

          Thank you, Per!

          I agree with you that the risk-weights are mindbogglingy stupid – nice way to put it too. As for your question about whether triple A or <BB- assets are more dangerous, I can't give a general answer. We both know the textbook answer, but I believe ratings can be very unreliable because of inadequate methodology, too much emphasis on ratings paid by the issuer, and because many securitisations are deliberately designed to game the rating system.

          On the second point, I am merely suggesting that they should get rid of RWA metrics altogether and focus on a high minimum leverage ratio – if we are not to have full free banking, which I would much prefer. Mixing up the two metrics is also a recipe for confusion and unintended consequences. I don't accept the view that high leverage ratios entail high social costs.

          • Kevin Dowd

            Oh dear! Its done it again :-)

        • Kevin Dowd

          Yeah George — duel at dawn mate!

  • Gary Anderson

    So, while I don't understand the ins and outs of risk leverage, it does seem like Basel allows the mispricing of risk, then forces lower ratios right when banks can't handle it. It is like mark to market or other central bank shenanigans. Misprice assets, cause a bubble, then take the bubble down by mark to market or by strict ratios, forcing the wealthy insiders to benefit disproportionately.

    Now Basel whatever wants to make all loans with no payments made on them for 90 days delinquent. That would really hurt Chinese banks. That is why China rejected Basel 2 and will not officially recognize any Basel shenanigans. Meanwhile in the west, our banks are crushed by regulation after the regulators mispriced the risk in the first place!

    And we wonder why people hate central banks the way they do.

    • Kevin Dowd

      Thank you Gary!
      On your first para, the way I would put it is that Basel encourages the mispricing of risk and the decapitalization of the banking system, despite ostensibly seeking to maintain banks' capital standards.
      On the second, my impression is that the Chinese banking system is a disaster with inadequate provision, accounts, capital, etc. – and yes, I agree our banks are crushed by pointless regulations that merely impede recovery.