Print Friendly and PDF

The Bear's Lair: The ultimate in foolish leverage

April 7, 2014

The Financial Times revealed this week that trades in index credit default swap (CDS) options had managed to avoid being listed on exchanges, with all the transparency requirements that brings, instead being allowed to continue trading on an over-the-counter basis.
The amount outstanding is relatively small in relation to the $25 trillion of CDS outstanding, but lack of transparency is likely to hide a deep underlying problem. I had thought that CDS themselves represented the ultimate in unmanageability by conventional risk management. But index CDS swaptions are worse, being even more leveraged and hence even more liable to excessively large tail risks that can crater the world's banking system.

Let's start with a little background, for those who never read our 2010 book, "Alchemists of Loss," or who have forgotten it. CDS were invented in the 1990s as a way to hedge/bet on credit risk. As an instrument, they have a number of problems, one of which (significant for these new gambling chips) being that there's really no good way to determine how much the thing will pay off in a bankruptcy. The CDS bankruptcy "auction" in which a few million dollars' worth of defaulted debt is put up for auction, to determine the price of instruments worth billions, is far too easily gameable. That's why, when swap market practitioners like myself had looked at the possibility of CDS in the 1980s, we had decided there was no practical way to create a sound product.

In managing CDS positions, the biggest problem is their huge amount of embedded leverage. A typical CDS will sell for perhaps 70 cents annual premium for $100 of 5-year credit insured.  It is thus leveraged about 30 to 1 in one direction ($3.50 potentially gets you $100 if you're a buyer) but very little in the other, since even if the credit improves and the CDS value goes to $0.40 per annum you only make $1.50 on your $3.50 short position if you're a seller.

This makes managing CDS positions very difficult indeed. In particular, the normal Gaussian risk-management systems such as Value at Risk (VaR)are especially inaccurate when assessing a CDS portfolio. If you calculate the standard deviation of CDS prices, it will be some fraction of their $3.50 initial value, yet the downside risk is $100 – 40 or 50 standard deviations, an outcome that is simply impossible under VaR. Even an outcome of a Gaussian distribution that varies 20 standard deviations from the norm cannot happen in the life of several million universes.

This is even more true for CDS swaptions. Here the two options, call and put, have grotesquely different risk profiles. A CDS put option has a very moderate risk. As I said above, it is unlikely that a CDS price will decline by more than a moderate fraction of its initial price. Hence a CDS put will have only moderate upside and will be priced at a small fraction of the CDS's nominal principal amount.

A CDS call option, on the other hand, has the full $100 upside (minus the strike price) if the credit defaults totally during the life of the swaption. Hence, since it presumably sells for a fraction of the price of the CDS itself, it is even more leveraged than the CDS. Its risk profile is even more pathological than that of the CDS and its risk management, if done by VaR or some variant thereof, even more underrates its risk.

As I have discussed previously, traders benefit enormously by finding instruments where conventional risk management grossly understates the risk of the instrument.  They are able to trade much larger volumes of the instrument than they could if it was managed properly. Hence, if their normal position is the "right way round" they will be able to lock in a high probability of a larger-than average profit that will result in a larger-than average bonus.

Of course, there is a modest risk that the position will go hopelessly wrong and bankrupt their employer, or close to it. But as we saw in the London Whale case, the risk to the trader himself is generally modest in such cases. (The risk to his hapless supervisor, and possibly to the institution's risk management leaders, will be much greater, which is certainly unfair in the case of the risk managers, since they will benefit little from the trader's above-average profit in good times.) The London Whale took these kinds of risks, made large bonuses as a result and came close to wiping out J.P. Morgan. Had he been playing in CDS swaptions rather than CDS his risk-reward ratio would have been even more favorable for him and even more likely to be fatal for the bank.

An additional wrinkle is that the new upsurge in trading volume is in index CDS, in which the underlying credits represent a basket of different borrowers. As was disclosed in the London Whale case, index CDS have a large amount of basis risk, with "aged" index CDS trading at less favorable prices than new ones, and the underlying credits in each new tranche being different.

This has allowed the industry to keep CDS index options from being listed on the exchanges, claiming that each option is too individualized to be traded in a standardized form. Of course, in reality, if the indexes are to be representative of a range of credit risk, and the options are to be of any use in hedging index CDS risk, the industry would seek to standardize contracts. This could certainly be done for CDS swaptions, which are mostly short-dated, even if the underlying CDS indices, which exist for several years, were more difficult to standardize.

However, keeping CDS index swaptions off exchanges keeps the market suitably opaque, and allows trading profits to be correspondingly greater, as unsophisticated institutional investors overpay for instruments for which they can neither accurately calculate a value nor determine an accurate market price. Needless to say, the fact that institutional investors are paying too much for their index CDS swaptions does not significantly reduce the risk to the banks selling them.

According to the Financial Times, the amount of CDS swaptions outstanding has recently grown rapidly, currently fluctuating between some $373 billion and $863 billion from one week to the next. That's not enough to wipe out the banking system, and indeed is surprisingly small compared to the $25 trillion of CDS outstanding. However, we know from experience with CDS themselves and subprime mortgages in the run-up to the 2008 meltdown that once this kind of super-profitable business gets going, the outstanding volume spirals very quickly. The number of unsophisticated investors sucked in rises exponentially and the exposure of the system rises out of control. After all, AIG Financial Products ran up $440 billion in exposure to mortgage bond CDS—enough to bankrupt it over a period of only a couple of years to September 2008.

Unless someone knows of a worse financial instrument, CDS options are the most leveraged and the most inaccurately rated by risk management of all products out there at this time. Since the profitability of a product to traders is directly proportional to the level of undercounting of its risks in conventional risk management systems, it therefore follows that CDS options are the place where the next banking system collapse is most likely to be concentrated.

Regulators should take notice. But I'll bet they don't. They're too busy carefully assessing the risks on products that are not especially risky, or that caused the last crisis. As the Bernard Madoff experience proved, the capacity of regulators to understand risks that are novel and that have not previously caused a collapse is infinitesimal. Since CDS options do not yet have a criminal record, the regulators will presume them innocent. The average local bobby on the beat knows better.

As I have previously written, CDS and options on CDS should be managed using a Cauchy risk management model, based on a distribution discovered by the French arch-Royalist mathematician Baron Augustin-Louis Cauchy (1789-1857). To generate a Cauchy distribution, you stand a rifleman with an infinitely powerful rifle and no tendency to dizziness on a turntable and spin him round. Each time he comes to a halt at a random angle, he fires a shot. Next to him is an infinite wall. Half the shots will miss the wall altogether; the other half will hit the wall, and the bullet holes will over time (a very long time) form a Cauchy distribution. The distribution has an infinite standard deviation, which makes it exceptionally effective in capturing the pathological risk of credit default swaps and their options, far more so than the wimpy Gaussian. (Cauchy despised Gauss, regarding him as a sickening middle-class liberal.)

A risk management model based on a Cauchy distribution provides far longer risk "tails" than a conventional model and therefore captures the potential leveraged downside of the CDS option product. A bank using such a model would set position limits for the product at infinitesimal levels, making it totally uninteresting to traders in the world's major financial institutions who would be forced to find a less destructive line of work. But then, the London Whale affair surely showed us if nothing else that this kind of instrument should not be traded at all in institutions whose deposit liabilities benefit from a state-backed guarantee.

There's always the hope that the regulators will force the adoption of Cauchy risk models by the banking system before the crash comes. But I wouldn't risk any money on it.


Commentary Archive