The biggest weakness of Basel III

By Felix Salmon
September 15, 2010
Economist and the second at The American Scene, Noah Millman does an excellent job of explaining the biggest weakness with Basel III. Which also happens to be the biggest weakness with Basel II.

" data-share-img="" data-share="twitter,facebook,linkedin,reddit,google" data-share-count="true">

I’m unapologetically happy and optimistic about the outcome of the Basel III process, and I haven’t been impressed by most of its critics — until now. In two posts, the first at the Economist and the second at The American Scene, Noah Millman does an excellent job of explaining the biggest weakness with Basel III. Which also happens to be the biggest weakness with Basel II.

The problem is that while Basel II was a bold experiment which took a decade to put together and which even then never really got implemented, Basel III was much more of a rush job, and therefore could not be a soup-to-nuts reimagining of what a global macroprudential regulatory regime should look like. Even if that were a good idea.

Instead, Basel III is essentially a bold new layer built over the old Basel II architecture, in much the same way that early versions of Windows were layered on top of DOS. And just as early versions of Windows shared some of the weaknesses of DOS, do has Basel III inherited some of the problems of Basel II.

The main one is the whole concept of risk weighting: the idea that some assets are riskier than others, and that banks should hold more capital against risky assets (unsecured loans to people with a 550 credit rating, say) than they do against much safer assets, like loans to the US government.

That makes a certain amount of sense, but there are two main problems with it. For one thing, it’s backwards-looking: it reckons that the securities which have been risky in the past are the same as the securities which will be risky in the future. That obviously isn’t true. And secondly, it’s easy to game. Here’s Millman:

The consequence of this Basel II reform was to discourage banks from lending to risky enterprises, and to encourage the accumulation of apparently risk-free assets. This was a primary contributor to the structured finance craze, as securitisation was a way to “manufacture” apparently risk-free assets out of risky pools. What brought banks like Citigroup and Bank of America to their knees wasn’t direct exposure to sub-prime loans, but exposure to triple-A-rated debt backed by pools of such loans, debt which turned out not to be risk-free at all.

Since it did not change this risk-weighting, Basel III effectively doubles down on Basel II. Banks will need to hold more common equity than ever—against their risk-weighted assets. That massively increases the incentive to find low-risk-weight assets with some return, since these assets can be leveraged much more highly than risky assets. Unless I’ve missed something, lending to AA-rated sovereigns still carries a risk-weight of zero. So one result of Basel III could be to encourage banks to increase their lending to sovereigns at the margins of zero-risk-weight status. If that happens, anyone want to guess where the next crisis will crop up?

This is all absolutely true, but at the same time a sovereign default is always going to cause a banking crisis, no matter what kind of capital-adequacy rules are in place. Central banks can’t protect banks from sovereign default, and neither can banks themselves. If you’re a bank and the country you’re based in goes bust, then you’re going to go bust too.

But Millman’s broader point is spot-on:

Since taking any additional measurable risk is now stigmatized, the game becomes how to increase returns without increasing measurable risk…

Developments in banking regulation in the last decade, meanwhile, have turbocharged this process, and I’m increasingly convinced contributed mightily to the financial crisis. At the heart of the financial crisis, after all, was banks investing in highly-rated debt backed by lousy mortgages. But why did they hold so much of this debt? In part because they could plausibly argue that it was risk-free or nearly risk-free… If the exposure was classified as market risk rather than credit risk, the Basel II framework was based on Value-at-Risk, which showed very low volatility.

The big-picture point to take home is: the regulatory framework recommended by Basel II assumes that banks are in the best position to measure their own risks, and that a regulatory framework that aligns regulatory capital requirements with the risk being taken is to be desired. In other words, the regulatory framework was pushing banks hard in the direction they were already going: towards avoiding measurable risks and hence (since you still have to make money) into risks you can’t easily measure (or don’t know exist).

As Joe Nocera explains, the whole Value-at-Risk structure gives banks every incentive to push risk into the tails. And because tail risk can be ignored, banks then go on to embrace other mechanisms — like the Gaussian Copula Function — which essentially fatten the tails, stuffing them with ever more risk. There’s not much in Basel III which directly addresses this problem.

And there’s another weakness in Basel III, too. I was at a Manhattan Institute event this evening, at which Paul Singer of Elliott Associates stood up and declared that the big systemic losses which resulted from the bankruptcy of Lehman Brothers were not actually a function of any kind of monster hole in Lehman’s balance sheet. Instead, after Lehman declared bankruptcy, its enormous derivatives book needed to be unwound very quickly, and it was that unwind which caused something between $50 billion and $75 billion of losses, precipitating the worst months of the crisis.

Singer wasn’t saying that the derivatives book caused Lehman’s collapse — far from it. The book was actually pretty kosher. The cause of Lehman’s collapse was liquidity problems, and Basel III does a pretty good job tightening up the rules on bank liquidity. But if a bank with a big derivatives book does end up declaring bankruptcy, the effects can still be catastrophic. Again, that’s not something that can readily be addressed in the Basel III architecture — it requires instead a lot of detailed tinkering with bankruptcy laws. Still, the tail risk remains.

Ultimately, Basel III does a good job of reducing foreseeable risks, and, like any ex ante regulatory structure, it does a bad job of reducing unforeseeable risks. The problem is that as a result, banks are incentivized to load up on the kind of securities which can blow up in unforeseeable ways. I’m not sure that’s something that the Basel Committee could ever really address, but it’s worth remembering, all the same.

More From Felix Salmon
Post Felix
The Piketty pessimist
The most expensive lottery ticket in the world
The problems of HFT, Joe Stiglitz edition
Private equity math, Nuveen edition
Five explanations for Greece’s bond yield
8 comments so far


I haven’t seen any mention of off-balance sheet assets and Basel III. Would appreciate it if in a future article you can address this issue. Thanks.

Posted by Polycapitalist | Report as abusive

“Since it did not change this risk-weighting, Basel III effectively doubles down on Basel II. ”

Well, that depends what you mean. Basel III certainly keeps the backward looking risk weighting approach, but it did dramatically increase the risk weighting on the assets that brought down Citi et al (CDOs of ABS). And, probably more importantly in the long run, it removed the major source of arbitrage under Basel II, the disparity in treatment between securitisations in the trading book and the banking book. For the most part, the problem wasn’t just banks loading up on CDOs. It was banks loading up on illiquid CDOs in the trading book, where they didn’t even have to hold the small amount of capital they would have had to hold if the assets were in the banking book, only a small market risk charge. Now, securitisation positions in the trading book get effectively the same treatment as the banking book, and the market risk framework has been significantly tightened up. As a result, the average bank is going to be holding about 2-3 times as much capital against its total trading book as in the past, even before you take into account the basic capital ratios.

Posted by GingerYellow | Report as abusive

Basel ii and Basel iii are both inherently based on a mean variance assumptions of risk. This is deeply flawed. Mean variance is nothing more than naive trend following which tends to severely over estimate component co-variance stability.

The more complex the system and bucketing of risk, the more fun people will have gaming it. Basel iii will most likely make the CDO act of stuffing 10 lbs of crap into a 5 lb bag look like amateur hour. The bag will of course be tiered capital requirements.

The only thing we can hope is that Basel gets implemented and people have so little faith in it that they force each other to be more transparent and not just really on a single metric such as tiered capital ratios.

I am now convinced the regulatory solution has 2 answers. No international bank regulation in which case all participants are wary of each other and tread more carefully or a reporting regime so full of holes that all participants are wary of each other and tread more carefully knowing the system is bogus and to be gamed.

Posted by NickGogerty | Report as abusive

Felix, it is a bit rich for you to be saying that you have only just noticed this little problem with B3. Your own blog commenters have brought up the points you mention and more besides. This issue of risk weights has long been a hobby horse of Per Kurowski, and he left a comment on your blog just the other day, did he not?

Posted by Greycap | Report as abusive

I think that by “mean variance”, Nick Gogerty is referring to mean-variance; that is, a statistical model parameterized by mean, variance, and correlation. I’m not exactly sure what Nick’s beef was because he wasn’t very specific.

What I would say: the problem is *not* the lack of “fat tails.” It is easy to construct mathematical models with fat tails, but impossible to calibrate them. This should be obvious given that we can’t reliably calibrate the parameters we already have (I think this might have been Nick’s point, in which case we agree.)

The underlying difficulty is that the distributions of asset returns are not endogenous properties; they are functions of what people think about the assets, and opinions are highly unstable, non-linear, full of feedback loops, etc. One could say that assets don’t really have return distributions. The consequence of this is that any model of return distributions must be continually updated, if modeled returns are to resemble observable returns. And the consequence of updating is that risk measures derived from these models are strongly pro-cyclical. You just have to look at historical time series of asset vols and returns to see what I mean.

So there is still a powerful pro-cyclicality baked into the foundations of B3. The ad hoc measures like variable capital ratios duct-taped onto the structure won’t be enough to mitigate this; not by an order of magnitude.

Posted by Greycap | Report as abusive

Re the RW discussion, it seems to me BCBS fully realises its limitations hence the introduction of the Leverage Ratio which sort of aims to backstop any model/model input error. Interestingly to see how the off B/S items (incl. netting) will be factored into this measure. Off course, too low weightings will allow for gaming the system.

Re the Lehman, BCBS pushes for Central Counterparties (allowing low RW) which should limit losses when derivatives are closed-out through adequate collateral mngt processes and multilateral netting. Note BCBS has also proposed to take longer time windows and increased haircuts for bilateral agreements.

BTW I’m surprised by all the complexity BCBS raises about the Trading Book. Why couldn’t they have just introduced Holding Period Limits (for Equity and Debt like instruments), breach of which leads to capital surcharges or alternatively -and rigourously- full deduction? In this way, banks would take on illiquid assets more prudently and if not willing to hold them at punitative capital levels, sell them onwards. It would seem to me that the market would do its work on valuing such illiquid assets.

Posted by MJGW | Report as abusive

A few things. First I’m pretty sure nothing should be 0% risk weighted. Second, yes, risk weightings are going to continue to be a big part of the real problem. Third Basel III fails to address the issue of the hegemony of the ratings agency cartel.

On another note, I remain fairly convinced that there isn’t any way to completely remove the chance that banks blow up. Banks are completely unlike any other corporate structure – they are inherently levered and can therefore be wiped out by a nasty run on the bank, even if perfectly solvent. So what’s the answer? The minute we admit that TBTF exists, we have to charge for the privilege of being TBTF rather than try to regulate it out. Insurance premiums payable to a pre-funded disaster recovery fund, that are heavily geared to (true) leverage and spiral exponentially above 30x leverage sounds like a good start. Throw in additional premia based on the % of revenues from trading, the type of funding that props up the balance sheet and the liquidity of underlying assets, and Bob’s your uncle, we might have the beginnings of something that works.

Posted by drewiepe | Report as abusive

Okay, I might be throwing a spanner in the works here, but aren’t there more effective risk management methodologies than just straight historical VaR?

The guys at RiskMetrics have been warning us for ages, well before the current crisis, of the risk underlying purely historical measures (and yes, this is even before Taleb’s black swan hit the public lexicon). And yet, from a regulatory point of view, we still try and define our risk on a calendar, five year, ten year, line in the sand kind of way, when we know that the reason markets are good is because they can react on information immediately: not with a particular timeframe in sight, but with a ‘real’ profit mindset.

Certainly, when it comes to futures and forwards, we’ve understood that margining is an ongoing process that needs to be constantly attuned so why is it that banks can not be given the same review/monitoring/risk management outlook? (I wrote an article on this recently at

Thank you Felix for highlighting the issue of sovereign risk but even sovereign risk need not become systemic. If it is acknowledged that even your own country’s securities still require some level of hedging then there should be some scope to aver the worse. There should be a level of freedom in the market to split the political operational risk from the real intrinsic economic risk – if markets are not allowed to incentivise this kind of rational behaviour then we are leaving ourselves open to ongoing confrontation between markets and local spendthrifts government: we should be able to work together on the long run, not just against each other.

Tariq Scherer

Posted by tariqscherer | Report as abusive
Post Your Comment

We welcome comments that advance the story through relevant opinion, anecdotes, links and data. If you see a comment that you believe is irrelevant or inappropriate, you can flag it to our editors by using the report abuse links. Views expressed in the comments do not represent those of Reuters. For more information on our comment policy, see