As I suspected, CDS collateral has a lot of complexities that I didn’t get to last time round. So with many thanks to my commenters and also to David Felsenthal of Clifford Chance, here’s more: on questions of buyers posting collateral, whether rehypothecation is allowed, and whether all the collateral being posted is causing a major liquidity shock. As ever, the mistakes are all my own.
First, do buyers of credit protection ever have to post collateral? The easy but unhelpful answer to this and all questions is that it depends on what is negotiated in the CDS contract. But if you take a normal CDS issued under the ISDA Master Agreement with few modifications, then yes, in many cases the buyer of protection does need to post collateral if the contract moves against them.
In practice, however, this is rarely a big deal. For one thing, spreads have been moving out, not in — which means that buyers of protection have generally been making money, not losing it, on a mark-to-market basis. But also there’s a limit to how much a buyer of protection can lose: if there are "points" up front then that money has been paid already, so a buyer’s only liability from then on in is the insurance premiums they have to pay every six months. Those payments are senior unsecured liabilities of the buyer, but if the seller of protection is worried about them, then they can ask for collateral to secure the upcoming payment. Normally the collateral requirement doesn’t exceed one payment.
Most of the collateral being posted goes the other way: from the seller to the buyer. And the big question there is whether the buyer can use that collateral in the normal course of business — rehypothecation, it’s called — or whether it just needs to sit there in the bank account, untouched.
The ISDA Master Agreement allows it both ways: it’s basically a check-the-box thing. Brokers nearly always want to allow rehypothecation of the collateral which has been posted to them, but any client can ask for rehypothecation to be "turned off" very easily. The more rehypothecation that’s going on, the less of a liquidity problem that brokers will have when they’re posting and receiving collateral every day: if they allow rehypothecation, the same collateral can effectively be used on lots of different contracts at once. But we’re moving towards a world where the brokers’ clients are increasingly concerned about the brokers’ own credit risk and are therefore more wary of allowing rehypothecation.
Historically, brokers sometimes negotiated CDS contracts with their hedge-fund clients whereby the hedge funds would need to post collateral if the contracts moved against them, but they themselves did not need to post collateral unless their credit ratings were downgraded. At this point, however, pretty much everybody is posting collateral, with the exception of Berkshire Hathaway, partly because the brokers did get downgraded.
Because the brokers tend to prefer to allow rehypothecation, most CDS contracts do allow it. (I was wrong about that last time round.) That minimizes the liquidity problems associated with lots of collateral needing to be posted at once, but it also introduces the risk of exactly what we’re seeing with Lehman Europe: counterparties being unable to retrieve collateral from a bankrupt broker because it’s been rehypothecated many times over.
Moving forwards, brokers may well start quoting two different prices for CDS: one for a contract with rehypothecation allowed, and the other one with it banned. Less likely is another proposal which has been mooted: that of holding the collateral at a third-party custodian such as Bank of New York.
The one thing I’m still very unclear on is how much total collateral is tied up in one form or another right now — and whether that amount is big enough to cause serious liquidity problems in the financial system. Obviously the numbers for AIG on its own are enormous, partly because it only ever sold credit protection and never bought any. But outside AIG and other formerly triple-A insurers, is CDS collateral significantly decreasing the liquidity available to investment banks? My very vague gut feeling is that the answer is no, but I might well be wrong about that.
Update: Alea points to an ISDA press release from April putting total collateral in circulation at end-2007 at $2.1 trillion, up sharply from $1.3 trillion at end-2006. Yikes! It’s not clear whether there’s any double-counting going on thanks to rehypothecation; for liquidity’s sake, I hope there is.
Update 2: I’ve now received a copy of the actual ISDA report; in fact there’s even double-counting without rehypothecation:
The objective of the ISDA Margin Survey is to estimate the
importance of collateralization in the market and not simply to estimate the value of assets used as
collateral. The Survey therefore tracks the gross amount of collateral–defined as the sum of all
collateral delivered out and all collateral received in by Survey respondents–and does not adjust
for double counting of collateral assets. Double counting takes at least two forms. The first occurs
when one Survey respondent delivers collateral to or receives collateral from another respondent.
The collateral assets in this case are counted twice, once as received and once as delivered. The
second source of double-counting is collateral re-use–sometimes called rehypothecation–in which
collateral is delivered from one party to another, then delivered to a third party, and so on. A single
unit of re-used collateral may consequently be counted several times by the Survey as the collateral
progresses down the chain of parties re-using it. But because each re-use represents the securing of a
separate and distinct credit exposure between two parties, we believe it is valid to count the collateral
as many times as it is used. If in contrast the objective were simply to measure the value of assets
currently in use as collateral, it would then be necessary to adjust for double counting.