Streetwise Professor

April 8, 2019

CDS: A Parable About How Smart Contracts Can Be Pretty Dumb

Filed under: Blockchain,Derivatives,Economics,Exchanges,Regulation,Russia — cpirrong @ 7:04 pm

In my derivatives classes, here and abroad, I always start out by saying that another phrase for “derivative” is contingent claim. Derivatives have payoffs that are contingent on something. For most contracts–a garden variety futures or option, for example–the contingency is a price. The payoff on WTI futures is contingent on the price of WTI at contract expiration. Other contracts have contingencies related to events. A weather derivative, for instance, which pays off based on heating or cooling degree days, or snowfall, or some other weather variable. Or a contract that has a payoff contingent on an official government statistic, like natural gas or crude inventories.

Credit default swaps–CDS–are a hybrid. They have payoffs that are contingent on both an event (e.g., bankruptcy) and a price (the price of defaulted debt). Both contingencies have proved very problematic in practice, which is one reason why CDS have long been in such disrepute.

The price contingency has proved problematic in part for the same reason that CDS exist. If there were liquid, transparent markets for corporate debt, who would need CDS?: just short the debt if you want to short the credit (and hedge out the non-credit related interest rate risk). CDS were a way to trade credit without trading the (illiquid) underlying debt. But that means that determining the price of defaulted debt, and hence the payoff to a CDS, is not trivial.

To determine a price, market participants resorted to auctions. But the auctions were potentially prone to manipulation, a problem exacerbated by the illiquidity of bonds and the fact that many of them were locked up in portfolios: deliverable supply is therefore likely to be limited, exacerbating the manipulation problem.

ISDA, the industry organization that largely governs OTC derivatives, introduced some reforms to the auction process to mitigate these problems. But I emphasize “mitigate” is not the same as “solve.”

The event issue has been a bane of the CDS markets since their birth. For instance, the collapse of Russian bond prices and the devaluation of the Ruble in 1998 didn’t trigger CDS payments, because the technical default terms weren’t met. More recently, the big issue has been engineering technical defaults (e.g., “failure to pay events”) to trigger payoffs on CDS, even though the name is not in financial distress and is able to service its debt.

ISDA has again stepped in, and implemented some changes:

Specifically, International Swaps and Derivatives Association is proposing that failing to make a bond payment wouldn’t trigger a CDS payout if the reason for default wasn’t tied to some kind of financial stress. The plan earned initial backing from titans including Goldman Sachs Group Inc.JPMorgan Chase & Co.Apollo Global Management and Ares Management Corp.

“There must be a causal link between the non-payment and the deterioration in the creditworthiness or financial condition of the reference entity,” ISDA said in its document.

Well that sure clears things up, doesn’t it?

ISDA has been criticized because it has addressed just one problem, and left other potential ways of manipulating events unaddressed. But this just points out an inherent challenge in CDS. In the case Cargill v. Hardin, the 7th Circuit stated that “the techniques of manipulation are limited only by the ingenuity of man.” And that goes triple for CDS. Ingenious traders with ingenious lawyers will find new techniques to manipulate CDS, because of the inherently imprecise and varied nature of “credit events.”

CDS should be a cautionary tale for something else that has been the subject of much fascination–so called “smart contracts.” The CDS experience shows that many contracts are inherently incomplete. That is, it is impossible in advance to specify all the relevant contingencies, or do so with sufficient specificity and precision to make the contracts self-executing and free from ambiguity and interpretation.

Take the “must be a causal link between the non-payment and the deterioration in the creditworthiness or financial condition of the reference entity” language. Every one of those words is subject to interpretation, and most of the interpretations will be highly contingent on the specific factual circumstances, which are likely unique to every reference entity and every potential default.

This is not a process that can be automated, on a blockchain, or anywhere else. Such contracts require a governance structure and governance mechanisms that can interpret the contractual terms in light of the factual circumstances. Sometimes those can be provided by private parties, such as ISDA. But as ISDA shows with CDS, and as financial exchanges (e.g., the Chicago Board of Trade) have shown over the years in simpler contracts such as futures, those private governance systems can be fragile, and themselves subject to manipulation, pressure, and rent seeking. (Re exchanges, see my 1994 JLE paper on exchange self-regulation of manipulation, and my 1993 JLS paper on the successes and failures of commodity exchanges.)

Sometimes the courts govern how contracts are interpreted and implemented. But that’s an expensive process, and itself subject to Type I and Type II errors.

Meaning that it can be desirable to create contracts that have payoffs that are contingent on rather complex events–as a way of allocating the risk of such events more efficiently–but such contracts inherently involve higher transactions costs.

This is not to say that this is a justification for banning them, or sharply circumscribing their use. The parties to the contracts internalize many of the transactions costs (though arguably not all, given that there are collective action issues that I discussed 10 years ago). To the extent that they internalize the costs, the higher costs limit utility and constrain adoption.

But the basic point remains. Specifying precisely and interpreting accurately the contingencies in some contingent claims contracts is more expensive than in others. There are many types of contracts that offer potential benefits in terms of improved allocation of risk, but which cannot be automated. Trying to make such contracts smart is actually pretty dumb.

Print Friendly, PDF & Email


  1. Great post. Again.

    As someone who has been pretty involved in CDS for the past 10 years I am amazed that there isn’t more “creative” use of choosing your circumstance in order to take advantage of CDS that has already been written. e.g. See the recent LBO where selective provision of guarantees in order to threaten orphaning of existing CDS gave a highly motivated buyer base for the refi of the bridge debt.

    And on your LNG post. I remember looking at gas storage sites in the UK over a decade ago where LNG was on the horizon but the assumption was that it wouldn’t have that much impact on summer-winter spreads. As the saying has it: when you make an assumption you make an ass out of you and umption.

    Comment by isp001 — April 9, 2019 @ 2:15 am

  2. Great sum up of the recent default determination reforms in the synthetic credit market.
    I have to say I don’t understand what is the link with smart contracts!
    If the point is to prove that we need a law court then you don’t have to use such a complex example.
    Smart contracts have different meaning, as an executable and distributed software code you can just use ISDA credit event information as a market data (as you would use Libor rate as a market data).

    If someone wants to code the credit event determination process, then it’s impossible as described in the article. What if we create a new CDS contract that is more deterministic? Example: same pay-off as the current contract but the buyer can trigger the CDS anytime and get delivered the cheaper bond (or cash equivalent) of a specific bucket. Thus, Removing the event part and keeping the price part.

    Comment by mxcx — April 9, 2019 @ 7:35 am

  3. Now I’ve completed the finance module of my MBA I finally have some inkling about what the other half of your posts are on about. 🙂

    Comment by Tim Newman — April 9, 2019 @ 9:07 am

  4. Early attempts to code artificial intelligence taught us more about how little we know about how the human mind works than anything about computers. Similarly, attempts to code autonomous blockchain contracts are showing us how little we know about how markets work (how little we know about what we imagine we can design, and the anthropomorphic computers in 60’s Star Trek and Lost in Space now look quaint in retrospect. I suspect that smart contracts (and financial automation generally) will also prove to be very useful in ways that don’t resemble current markets, but current iterations mimic human institutions while ignoring the integrity, flexibility, and experience of market participants that make them work. The result is as naive as anthropomorphic depictions of computers ignoring the myriad of cognitive shortcuts and intuitive leaps that compromise human intelligence.

    Sometimes financial automation can start out in a working state, and then fails to scale when the money gets big enough to make gaming the metrics worthwhile. Pay-per-click-through on online ads worked reasonably well when it was small potatoes, but when the money got big, click-fraud arose on massive scale. User metrics on social media also started off relatively benign, until the stakes got so high that we now see massive data dredging, social influencers with armies of fake robotic followers, and so on.

    Comment by M. Rad. — April 9, 2019 @ 4:31 pm

  5. @Tim–Welcome to the Illuminati!

    Comment by cpirrong — April 9, 2019 @ 6:34 pm

  6. A chunk of that text is missing…

    (how little we know about what we imagine we can design,…
    and all that). Computers, of course, turned out to be useful for things that don’t resemble human intelligence
    …and the anthropomorphic computers in 60’s Star Trek and Lost in Space now look quaint in retrospect.

    I hope that is more intelligible.

    Comment by M. Rad. — April 9, 2019 @ 7:02 pm

RSS feed for comments on this post. TrackBack URI

Leave a comment

Powered by WordPress