The CDS Losses Arms Race Continues

There seems to be some kind of an arms race going on to see who can come up with the scariest number for total losses in the credit default swap (CDS) market. Remember Bill Gross with his $250 billion? Well now Andrea Cicione of BNP Paribas has trumped that with a "worst possible scenario" of $1.4 trillion in losses.

Cicione explicitly includes counterparty risk in his estimates:

A big bank, for example, may have sold protection in $120 million of contracts on a company and bought another $100 million in protection on the same company, leaving it with a much smaller net loss of $20 million.
But if a counterparty fails to pay on some contracts, Cicione said, then those contract losses are no longer just notional but add to an investor's net losses.

Now it't true that there is non-trivial counterparty risk in the CDS market. But it's also true that even though CDSs aren't traded on any exchange, they're still governed by something called the ISDA Master Agreement, which allows for "cross-transaction payment netting". If Andy sells $100 million of protection to Bill who then sells $100 million of protection to Carl and so on all the way to Greg who sells $100 million of protection to Andy, then any default should at that point cause no problems at all, since everybody's netted out to zero. And since there are no payments, there's no counterparty risk.

(Update: Steve Waldman and Andrew Clavell, in the comments, say that there is counterparty risk in this scenario, and that cross-transaction netting won't help. They know more about this stuff than I do, I'll take their word for it.)

What's more, the CDS market also allows something called "close-out netting", which helps significantly in the event that a counterparty does go bust. Let's say Andy sold $100 million of protection to Bill and then Bill sold $100 million of protection back to Andy. In that case, Andy is safe even if Bill goes bankrupt. Yes, Bill owes Andy $100 million and can't pay it. But Andy also owes Bill $100 million, and can simply net that out unilaterally.

So while the likes of Gross and Cicione are worried about gross exposure, in reality the markets have managed to build a structure where the real risk is only to net exposure. As the Office of the Comptroller of the Currency explains in its most recent Quarterly Report on Bank Derivatives Activities,

For a portfolio of contracts with a single counterparty where the bank has a legally enforceable bilateral netting
agreement, contracts with negative values may be used to offset contracts with positive values. This process
generates a "net" current credit exposure.

According to the OCC, this net exposure for US banks was $256 billion at the end of the third quarter. Clearly actual losses are going to be a tiny fraction of that, since for every counterparty who loses money there'll be another who's making money, and only the ones losing money will go bust. What's more, most counterparties are hedged: there aren't all that many entities who are writing lots of protection and not buying any.

Oh, and did I mention? The $1.4 trillion of losses in Cicione's worst-case scenario are notional. The actual amount of money that counterparties might fail to pay, says Cicione, is about one tenth of that sum: $150 billion. And that's assuming some extremely pessimistic numbers:

Cicione assumed a worst-case default rate of 4 percent and a 25 percent recovery rate. "These are aggressive estimates, the worst that can happen," he added.

Those assumptions compare to Gross's, of a 1.25% default rate and a 50% recovery rate. If you use Gross's assumptions instead, then chances are there wouldn't be any counterparty defaults at all. But even if there were, they would be a maximum of about $30 billion or so - if you divide Cicione's $150 billion first by 3.2 to account for the lower default rate, and then take two thirds of that figure to account for the higher recovery rate.

In other words, Cicione's numbers could turn out to be more optimistic, not more pessimistic, than Gross's - although it's unclear what numbers Cicione puts on losses which don't cause counterparty defaults. One thing we know for sure, though, is that his "notional" losses will always be much higher than actual gross losses, thanks to those netting arrangements.

(Via Campbell)