Books by Usha Alexander

  • A lone woman travels fearlessly into the jungle to confront the enemy. She holds the fate of an entire world in her hands.

  • When Craig Olsen returns to Idaho to say goodbye to his dying uncle, who raised him, he comes face to face with matters he can no longer evade.

  • "A suspenseful narrative weaves the stories and secrets of two generations into one seamless drama ... a worthy literary journey." —Kirkus Discoveries

Namit Arora's Photography

Selected Videos


  • StatCounter

« Jonathan Haidt at TED | Main | The American Electorate, 2008 »

November 01, 2008


The size of the CDS market is staggering no doubt, but should be understood carefully. First, it is only the sum of the nominal value of the contracts. Secondly, it involves over-counting, since many firms are involved heavily both in buying and selling CDS contracts on the same underlying entities, thereby having much lower net exposure. In other words, it means that the potential losses could be much much lower.

For example, the nominal value of CDS contracts hinging on default by Lehman Brothers was estimated to be $400 billion. The>recent clearing among the counterparties involved amounted to net payments of $5.2 billion. This despite the fact that sellers of protection had to pay 91 cents to the dollar, an unexpectedly high amount.

Assuming the same ratio of clearing amounts to nominal value, the CDS market implies a net exposure worldwide of $715 billion. That is a large amount, but not financial armageddon.

The confused state of affairs is a result of this being an over-the-counter (OTC) market. One immediate and sensible thing to do is to convert it to an exchange-traded market. The exposures would then be immediately apparent and systemic risks can be seen and addressed early on.

Great points. It's good to have an ex-physicist in the financial industry as a co-blogger. Did you see this coming and, if so, for how long?

I had realized that due to multiple buy and sell CDS contracts on the same underlying asset, the net collective exposure is a fraction of their total $55 trillion face value (I do something similar with call / put spreads on individual stocks). The Lehman CDS settlement is encouraging in its less-than-expected magnitude. But with so many gurus also surprised by it, I wonder if this was a relatively lucky case.

As I see it, here are some systemic problems with CDS, why I think the exposure may be far higher than your estimate, and some questions:

a) The textbook goal of CDS is to redistribute risk across multiple willing parties. This is all well and good except that it also encourages irresponsibility. As a lender, if I can redistribute my risk from a subprime loan across nine other parties, I can make up to ten such loans for the same total exposure to me, even as the collective risk to the system from my actions is now up to ten times higher.

b) In the short-term, the damage from (a) can create system-wide credit crunches like we just saw, requiring bailouts and drawing taxpayers into a game of roulette they did not choose. During market shocks, the leverage effect in these instruments amplifies both the loss and the gain for each player. Depending on where the chips fall and the assumptions used in a player's risk model, it can quickly wipe out a whole lotta players (like Lehman). Furthermore, CDS contracts now extend to corporate and municipal bonds too (how much of that is subprime?). If the recession deepens, won't the subprime mortgage crisis spiral down and spill over, making future bailouts much larger?

c) Moving CDS to an exchange traded market will certainly improve transparency (what about past CDS trades though? Most of them will be around for 4 to 5 years). But will it also make the exposure to the system "immediately apparent"? The notable point critics like Buffet and Soros make about CDS is that we do not fully understand how they work (for e.g., they're not based on risk from independent random events as in the case of, say, personal auto insurance, but on events that influence each other in complex ways).

In my previous job, I worked for a while on building mathematical models for default and bankruptcy by companies. This also involved some empirical work on the CDS market. The one standout thing we learned was that our model, based on what is known as the>Merton model, predicted a much higher default probability for financial firms than the CDS market was implying. We pointed this out to our clients, but did not explicitly claim that disaster was looming. While I couldn't have predicted the collapse of a specific firm, I did realize that the financial firms were very highly leveraged without any apparent consequences. This by itself was an unsustainable situation.

A few comments / answers to your questions:

a). A really useful thing to remember is that a CDS contract is an insurance contract. Thus, by itself, it does not involve multiple parties. I think that Soros, Buffett et al. are more worried about a structure known as a>Collateralized Debt Obligation (CDO). This is a structure in which:
1. many financial instruments are pooled together (thus aggregating their risks),
2. the pool is broken up into different tranches which have different risks, and
3. securities representing claims to the various tranches of the structure are sold (allowing buyers to choose their risk exposure).
This is the mechanism for spreading the risk among multiple parties. We must remember however, that no new risk is created by the spreading. Indeed, if you "make up to ten such loans", the total risk is higher than making only one such loan, but the CDO structure does not cause the increase. It does allow the loan originators to abandon all caution, but then, the sponsors of the CDO structures still have to find ultimate buyers of CDO securities. The willingness among investors (hedge funds, insurance firms, mutual funds) to buy subprime CDO securities, aided by the high ratings provided by the rating agencies, were at the root of the systemic problem.

b.) Your general comments are correct, but perhaps you are thinking about CDO structures based on subprime mortgages. CDS contracts were initially offered only on corporate and sovereign debt, but are being extended. A small note on terminology: 'subprime' usually refers to mortgages. Other types of debt with weak credit ratings are called 'below investment grade debt'. In the corporate CDS market, contracts on investment-grade firms had much higher volume a couple of years ago, when I last looked at the data.

Indeed, if the recession deepens, defaults among borrowers with higher credit ratings will increase. Whether that will create further crises that require further bailouts is difficult to predict, but my instinct is that it is unlikely.

c.) Futures exchanges give us good models for how a CDS exchange may work. First, all participants have to post margin collateral, which controls the number of contracts that a participant can be a counterparty to. Secondly, because the exchange is a clearing house, all contracts are netted out and the net exposure of each participant can be known at the end of every trading day. So while it may not be public knowledge, the exchange (and regulators) will know exactly how much exposure each financial institution has. As for already existing CDS contracts, governments can make laws requring registration at the exchange.

Finally, as I have mentioned before, I think Buffett and Soros are referring to CDOs, which are indeed difficult to understand or model scientifically. If the goal is to avoid systemic risk, regulators and central banks can easily impose limits on the extent of exposure that banks and financial institutions can have to CDO securities.

[Note from Namit, 1/19/09: I noticed that Typepad dropped my response to vp above, probably during a software upgrade. Fortunately, I have a copy of my response which I have inserted below.]

The CDO market is ~$1 trillion, the CDS market is ~$55 trillion (2007 data). Not sure how CDO can be more worrisome than CDS. I'm no expert but isn't CDO more often a packaging technique (pool/tranche based as you write) that relies on CDS to distribute risk? For instance, "synthetic CDOs gain credit exposure to a portfolio of fixed income assets without owning those assets through the use of CDS."

Further, while a CDS contract is bilateral, the risk from a given CDS can also be spread among multiple parties through the technique of netting:

Chains of CDS transactions can arise from a practice known as "netting". Here, company A may buy a CDS from company B with a certain annual "premium", say 2%. If the condition of the reference company worsens, the risk premium will rise, so company B can sell a CDS to company C with a premium of say, 5%, and pocket the 3% difference. However, if the reference company defaults, company B might not have the assets on hand to make good on the contract. It depends on its contract with company A to provide a large payout, which it then passes along to company C. The problem lies if one of the companies in the chain fails, creating a "domino effect" of losses. For example, if company A fails, company B will default on its CDS contract to company C, possibly resulting in bankruptcy, and company C will potentially experience a large loss due to the failure to receive compensation for the bad debt it held from the reference company. Even worse, because CDS contracts are private, company C will not know that its fate is tied to company A; it is only doing business with company B. (source)

It seems to me that netting also increases the systemic risk (while reducing exposure for each party) because only one party in the chain now has to default, a higher probability event.

Finally, in your last para, you state that CDO is the risky instrument, "difficult to understand or model scientifically". Is your implication that CDS does not share this problem? Is there any evidence at all that we can scientifically model the probabilities of economic events that are not sufficiently independent of other economic events (as in the case of, say, auto accidents), or that their interdependence is amenable to scientific description (how to account for human greed, for example, which can alters credit ratings, say). Do you think it likely that these are instead emergent properties of the system, and scientists may be just fooling/enriching themselves?

You are right about one thing: CDO structures with CDS instruments are indeed common and one order more complex than so-called "cash" CDOs, which contain straightforward debt securities.

The Wikipedia description of "netting" is a bit confused, so let us try and clean it up here. Each CDS contract has a protection seller and a buyer. If party A sells protection on the one hand to B and buys protection on the other from C, the difference in premiums can indeed be a source of profit. The risk here is counterparty risk, which is the risk that a counterparty to a contract will fail to deliver on its obligations. Counterparty risk can indeed propagate throughout the financial system.

At this point, it is best to clarify what we mean by risk. When referring to the CDO structure, I used the word "risk" to mean the chance of loss coming from defaults in the underlying pool of instruments. This will remain the same, no matter whether all the instruments in the pool are held by one party or many. What you are thinking of however, is the further risk that these losses will cause the failure of the buyers of CDO securities. For one CDO structure, the total exposure is clearly divided up by spreading. However, if many firms systemwide decide to tank up on CDO securities and defaults within the structures are high, this may indeed cause many failures.

This however, is not a problem with CDOs, per se. Imagine, for example, that many banks in a certain economy decide to buy a lot of price-risky assets (say paintings) during a boom period. When the boom turns to bust, they all lose a lot of money and face failure. [If this scenario seems unreal, replace paintings with real estate and you can come up with real-life examples]. This systemic issue is normally tackled by regulators using bank examiners and the like, who can (and very often do) specify what kinds of assets banks can buy and the extent of their exposures. The current failure of the regulatory mechanism, at its root, lies in the blind faith put in the ratings attached to risky instruments by ratings agencies such as S&P and Moody's.

At the level of the individual firm, failure can only be prevented by prudent management. In the recent crisis, one would have thought that self-interest would have prevented individual firms from being overexposed, but apparently greed overcame fear.

As I said in my previous comment, systemic risk can be mitigated by making these instruments exchange based, with margin requirements, in analogy with futures exchanges. Since each contract would require margin collateral, it would clearly curtail the chain-linking of contracts, limiting systemic risk. Regulators would also be able to get a clear view of the exposures of institutions under their control, directly from the exchange.

Regarding your questions about modeling: I was referring to pricing models. The goal of these models is to come up with the price that should be paid for an instrument, given the terms and conditions. For a CDS contract, this is the premium paid to the protection seller. The premium depends on the probability that the underlying entity will default. There are at least a couple of reasonable approaches to this, which give good agreement with results (one of these was the Merton model I referred to in my comment). For CDOs, however, it is quite tough to come up with a good model because of the interactions of multiple random variables. There are fancy mathematical structures to describe interdependence called>copulas, but it is not clear that they are necessarily the right tools for the job.

Finally, I think we should realize that most mathematical modeling in finance is limited in scope, just as it is in physics and engineering. The real world is often too complicated, and the key to successful modeling is abstraction and idealization. So, even if certain properties are emergent, one might be able to build descriptive models without necessarily aiming to describe how emergent properties emerge. I don't think people are fooling/enriching themselves in trying to model complex financial instruments - they are just trying to get a handle on how these things behave.

> Finally, I think we should realize that most mathematical modeling in finance is limited in scope, just as it is in physics ...

Whoa. The dismal science on par with sublime physics? As McCain might have said, the world of money is corrupting you, my friend.

On the other hand, the much needed Obama win today might not have been possible without these toxic, ill-understood, and unregulated credit derivatives that have helped make the economy the #1 issue for the electorate.

Verify your Comment

Previewing your Comment

This is only a preview. Your comment has not yet been posted.

Your comment could not be posted. Error type:
Your comment has been saved. Comments are moderated and will not appear until approved by the author. Post another comment

The letters and numbers you entered did not match the image. Please try again.

As a final step before posting your comment, enter the letters and numbers you see in the image below. This prevents automated programs from posting comments.

Having trouble reading this image? View an alternate.


Post a comment

Comments are moderated, and will not appear until the author has approved them.

Your Information

(Name is required. Email address will not be displayed with the comment.)

Primary Editors

New Book by Namit Arora

  • The Lottery of Birth reveals Namit Arora to be one of our finest critics. In a raucous public sphere marked by blame and recrimination, these essays announce a bracing sensibility, as compassionate as it is curious, intelligent and nuanced.” —Pankaj Mishra

Shunya Website

Namit wins 3QD Arts & Literature Prize 2011

Namit Arora's India Photo Archive