It has been widely reported that the ISDA settlement auction for Lehman Brothers' credit default swaps comes up today, Friday, October 10th.
The Financial Times reported on October 1 that Lehman's bonds are trading at 15-19% of face value, and immediately concluded that issuers will wind up on the hook for 81-85% of the notional value of Lehman CDS. This implicitly assumes that the auction will produce the same price as the post-default secondary market. It will be very interesting to see how that assumption holds up.
There hasn't been much discussion in the press on the counter-intuitive results of the Fannie/Freddie CDS settlement auction. Most commentators seem to think that it's just great that Fannie's senior CDS settled at 91.51 and its subordinated at 99.9, since the CDS writers won't need to pay out much cash (in comparison to a scenario in which Fannie had been permitted to default on its bonds). Felix Salmon rightly calls this a "topsy-turvy" outcome, and quotes Alea's explanation in terms of the cheapest-to-deliver effect.
The cheapest-to-deliver effect, in which the bids in a settlement auction tend to reflect the market price of the cheapest bond in the deliverable class, is of course real. But a closer look at the final auction results shows that it is not the correct explanation for the senior/subordinated price discrepancy. The inside market phase of the auction produced almost identical prices for Fannie's senior and subordinated debt (92.4 and 92.65 cents on the dollar respectively), and likewise for Freddie (93.75 and 93.8 respectively) - all presumably more or less consistent with the cheapest-to-deliver effect.
Where the senior and subordinated auction results differed is in the limit order phase, which was distinguished by the fact that for the first time in history the net open interest after physical settlement was to buy bonds rather than to sell them. The net open interest in the senior debt was quite small; the effect of the limit order phase was to raise the settlement price for Freddie senior CDS by 0.25% and to lower the price for Fannie senior CDS by 0.89%. (This paradoxical effect, a price shift in the "wrong direction" relative to the net open interest, is an artifact of the auction procedure and is clamped to a maximum of 1% by the protocol.) But the net open interest to buy subordinated debt for settlement purposes was so large as to climb up the limit order sequence all the way to 99.9% for Fannie and 98% for Freddie.
The upside-down settlement prices for Fannie's and Freddie's CDS - in Fannie's case, 849 basis points for senior CDS last traded at 38, and 10 basis points for subordinated CDS last traded at 218 - was therefore a direct result of a CDS volume (even after bilateral netting and pre-auction private settlement) that greatly exceeded the free float in the deliverable subordinate bonds. This has three rather alarming consequences for the CDS market.
1) When a heavily traded CDS actually suffers a credit event, there is likely to be a large net open interest to buy bonds for physical settlement. In a real bankruptcy event with substantial bond impairment, the cash settlement price is therefore likely to be much less than a real bondholder's loss on the same notional amount of bonds. The CDS no longer serves its "primary" function of hedging default risk on the bond it is nominally written to "insure".
2) Specialists in the CDS market, who have the wherewithal and motivation to simulate the outcome of CDS settlement auctions, hold a great information advantage over general investors, especially during the 30-day period between a credit event and the settlement auction that it triggers. This applies with especial force to major auction participants, who may well have the knowledge and motivation to game the auction protocol itself.
3) Worst of all: the main role of CDS in today's market is as an ingredient in "synthetic securities" whose high credit ratings reflect the rater's confidence that the risks intrinsic to their core assets have been properly hedged. The models used in designing and rating these structured vehicles presumably take into account the "cheapest to deliver" effect - insofar as this is possible without knowledge of future bond issues that may become eligible for delivery under the master settlement protocol. They probably do not (yet) take into account the "net open interest to buy" effect, which is in any case much less straightforward to model (since it depends on opaque variables like the ratio of notional CDS to free float in deliverable bond, as well as imponderables like the proportion of gross outstanding CDS that may be settled privately prior to the auction). These securities are therefore probably nowhere near as risk-neutral in practice as the existing models suggest.
This brings us to the Lehman auction. There are some $138 billion of senior bonds and $17 billion of subordinated bonds oustanding (according to Lehman's Chapter 11 filing); according to the ISDA's lists of deliverables and non-deliverables, all of the subordinate bonds and a substantial fraction of the senior bonds are non-deliverable under the settlement protocol, and one may expect that many others are locked up in structured investment vehicles, further reducing those available in principle for physical settlement.
According to Citi via the FT, there is some $400 billion of notional volume in Lehman CDS outstanding, perhaps 80% of the volume of CDS written on Fannie and Freddie's $1.4 trillion debt. However, there is a critical difference: according to Reuters, less than $15 billion of Fannie's and $5 billion of Freddie's bonds were subordinated. (This scarcity, together with the absence of any zero coupon subordinated debt that would actually be profitable to sell given the cheapest-to-deliver effect, explains why there were lots of bids but no offers in the physical settlement phase of the Fannie/Freddie auction.) I have no idea what fraction of the Fannie/Freddie CDS was written on the subordinated debt - perhaps I should ask Michael Hampden-Turner, who seems to be the FT's preferred source on these matters - but it probably greatly exceeded the free float in the relevant bonds.
In sum: I'm not going to pretend to predict the outcome of the Lehman auction, not just because the necessary data is unobtainable but because I'm an amateur analyst who may not have the story straight. But I am confident in saying that CDS pricing is even less transparent than it seems at first glance. Setting aside the counterparty risk, the liquidity risk, and the litigation risk (due to absurdly complex CDS contracts), the cash settlement procedure can and does set prices on swaps that are wildly different from their ostensible value as "get out of default free" cards to be paired with some specific physical bond.
The classic trouble with derivatives - as generation after generation of investors has learned the hard way - is that their price has the intended mathematical relationship to the underlying security, right up until it doesn't. The one circumstance in which the value of a credit default swap is certain not to match the risk it is supposed to counterbalance is when there is really a default. Add to this the potential for the CDS tail to wag the creditworthiness dog, and they start to look like not such a good idea.
Where things really get nasty is when risky debt is packaged with derivatives in order to synthesize an asset that appears risk-free. And where they get catastrophic is when these synthetic assets are rated AAA, purchased by entities restricted (by regulators, capital adequacy rules, or prospectuses) to holding only AAA-rated assets, and impaired or de-rated simultaneously with all the other AAA-rated synthetic assets built from the same blueprint. But that's a topic for another article.
To anyone who is watching closely, the Fannie/Freddie auction has already demonstrated that credit default swaps trade at the prices they do, not because investors understand their likely value after a foreseeable "credit event", but simply because the arbitrageurs' models say they should. When commercial paper markets price (or withhold) credit based on quoted CDS prices, they validate those models, and in the process they can easily "validate" a company like Lehman right into the ground. To add insult to injury, it seems likely that they can't be relied upon to do what they're supposed to after an honest-to-God bankruptcy. It will be a delicious irony if it is Lehman's settlement auction that brings this likelihood home to the market at large.