Regulators have been largely supportive of the credit default swap market’s efforts to move all standard CDS (a.k.a. “vanilla” CDS) contracts onto a central counterparty (CCP). Within a relatively short amount of time, the CDS market garnered the support of both the SEC and the Federal Reserve, setup shop, and executed CDS transactions totaling $71 billion in notional amount on ICE Trust LLC (ICE), the first of what could be a handful of CCPs. Given that the CDS market is still serving time on the pillory, this kind regulatory largess seems to imply that a CCP must undoubtedly be a good thing. However, not everyone is convinced. In a recent paper, Darrell Duffie and Haoxiang Zhu of Standford University described the state of affairs with and without a CCP in mathematical terms, proposed a measure of efficiency which they use as a proxy for counterparty risk, applied this measure to each state of affairs, and came to conclusions that are surprising, but not counterintuitive once you take a moment to consider the trade-offs between a distributed dealer system and a CCP.
There are two central points in their paper: (1) the benefits of having a CCP compared to not having a CCP (the distributed dealer system) is a function of the number of dealers (i.e., the more dealers, the more the system benefits from having a CCP) and that the current number of CDS dealers might be too small to realize any benefits from a CCP; and (2) even if having a single CCP is more efficient than not having one, having more than one CCP is never more efficient than having none at all. The assumption underlying these two points is that having a CCP as opposed to not having a CCP is in essence a choice between (i) multilateral netting across a single asset class and (ii) bilateral netting across multiple asset classes. After a bit of back and forth with Duffie, I agree that this is the case, but only in the short term. That is, there is nothing about a CCP that precludes netting across asset classes, even if the current model doesn’t facilitate it. I also disagree with the two central conclusions, mostly due to shortcomings of the model that the authors acknowledge in their paper. Specifically, they acknowledge that their model doesn’t deal with so called “knock-on effects,” or simply put, how one dealer default can lead to another. I also disagree with how Duffie and Zhu measure the benefits of netting, but we’ll spare our brains the heavy lifting and focus on the more practical issues. In this article, I’ll focus on explaining why a CCP does not preclude netting across asset classes. In a follow up article, I’ll explain how a CCP mitigates counterparty risk by facilitating trade compression.
What Is A Central Counterparty?
Rather than provide a one line, academic definition, I’ll proceed gradually, and by way of example. For starters, a CCP is not an exchange. Even with a CCP, inter-dealer trades will still be entered into between dealers, and price discovery will still take place across dealers. That is, dealers will still trade with each other, at least initially. After two dealers enter into a CDS trade, they will transfer, or “novate” their positions to the CCP. For example, in the prototypical CCP transaction, if Dealer A sells protection to Dealer B, each dealer would then novate its position to the CCP. After the novation, the state of affairs is such that A sells protection to the CCP and the CCP sells protection to B. When payments are made on the CDS, they get made to the CCP and then passed on to the parties.
At first blush, it might seem like all we’ve done is throw an extraneous and useless 3rd party into the transaction that does little more than operate as a conduit. However, this is not the case. The CCP is a distinct entity that has an interest in its own survival, and since the CCP is now liable to both dealers, it has an interest in being well-capitalized. While individual dealers also have an interest in being well-capitalized, they are unable to determine whether the CDS market as a whole is well-capitalized. And even if they were able to determine whether or not that is the case, each individual dealer has no power to convince others to increase the capital allocated to their positions. By centralizing all of the trade information into one entity and giving that entity control over the levels of collateral that dealers must post, we have created a method through which we can better ensure the adequate capitalization of the entire CDS market. For example, if one dealer is unable to fully collateralize its positions, the CCP can draw on its own capital and the capital of the other dealers to make up for the shortage. Without a CCP, that shortage of collateral would probably trigger events of default, which could “knock-on,” or cascade through the market. But as always, there’s no free lunch, and despite all of this upside, we have also concentrated the risk of counterparty failure.
Payment Netting And Payment Settlement
As Duffie and Zhu note, dealers net their payments and collateral across different types of vanilla swaps. So, for example, if Dealer A owed Dealer B $5 under a vanilla interest rate swap on some payment date, while Dealer B owed Dealer A $3 under a vanilla CDS, A would simply pay B $2. This type of bilateral netting across asset classes reduces the risk that dealers will default when compared to not having such netting because it reduces the total amount of cash that is needed to meet payment obligations. Duffie and Zhu argue that because CCPs segregate vanilla CDS trades from all other types of vanilla swaps, we miss the opportunity to net across asset classes. This is not necessarily the case.
First, we need to distinguish between the entities that are liable for the trades (i.e., the dealers and the CCP) and the entities that actually process the trade information and payments. In the case of ICE, trade information is processed by both TCC and DTCC. After that trade information gets processed, DTCC submits payment instructions to CLS Bank, which actually handles the payments. So, the CCP itself is not able to net payments across various asset classes, since it only handles CDS trades and moreover doesn’t actually settle the payments. However, DTCC will submit payment instructions to a settlement agent, CLS Bank, that actually handles the cash payments. CLS Bank could also receive payment instructions for other asset classes from the dealers (i.e., payment instructions on interest rate swaps, etc.) and net the payment instructions from the DTCC against the payment instructions from the dealers’ other activities. In fact, CLS Bank is also a leading settlement agent in the FX market.
In short, netting CDS contracts first does not preclude you from netting against other swaps later, especially since trading and settlement are handled by distinct entities.
Originally published at Derivative Dribble and reproduced here with the author’s permission.