Since 2009 credit event auctions that determine credit default swap payouts have been hardwired into the operations of the market. Previously, auctions were arranged on a more ad hoc, voluntary basis.
Potentially billions hinge on the outcomes of this automatic mechanism. By eliminating voluntary participation, and by virtue of dealing with something as complicated as debt, both credit events and the auctions that follow have been subject to the occasional controversy.
Auctions in 2011 included a string of Irish banks, Dynegy, Victor Company of Japan, Seat Pagine Gialle, and PMI Group. Rounding off the year was AMR Group, the parent company of American Airlines.
In previous posts, FT Alphaville explained why such auctions came to exist and how they served to facilitate the huge amount of growth in the market, particularly in tradable indices like the Markit iTraxx and CDX series.
Here we discuss auction mechanics.
Bring forth, the penguin!
We’re getting awfully fond of this little guy. He’s going to illustrate what physical settlement looks like, as this is the original way that CDS were settled.
The penguin, who was holding bonds and using CDS to hedge them, delivers the bonds of the defaulted issuer to the protection seller. In return, the penguin receives the par value of the bonds, compensating Happy Feet for his losses. The CDS contract is torn up at that point.
There is no more physical settlement since the 2009 Big Bang Protocol that all industry participants signed up to.
CDS now only end their lives, post credit event, by way of cash settlement, and here’s what it looks like:
The final outcome is the same in both diagrammes.
The second one illustrates a feature that makes the credit derivative market attractive — there’s no need to own the underlying bond, or even have any other exposure to the underlying. (A possible exception to that will be if the proposed ban on naked shorting of European sovereigns comes into force in November, as planned.)
Auctions answer the question: how much cash needs to be transferred to the buyer to settle the contract? They do this by facilitating trading in some of the reference entity’s debt, seeing at what price it clears. The clearing price of the debt is termed the “final price”, and cash settlement for CDS is dictated by one minus the final price (the loss experienced on the debt).
Signing up to the Big Bang Protocol, and all participants did, means that you must cash settle and it must be at the amount determined in the auction.
The auction process has two stages and both are made public.
The terms of each auction are laid out by Isda, a trade body, in consultation with its membership. The auctions are run jointly by Markit and Creditex, and participating dealers have to be prepared to sell or buy a minimum amount of the bonds being auctioned. The dealers may be acting on their own account, on behalf of their clients, or both.
Stage one
In the first stage of the auction, all the participating dealers have to submit prices for a pre-approved pool of bonds. The bonds of a defaulted entity can be very illiquid. On the one hand, the auction serves to at least lay out a uniform procedure for dealing with that, as the process dictates how CDS payouts will be established. On the other, the results can be rather curious sometimes.
Regardless, here’s what it looked like when dealers submitted prices for the pool of Dynegy bonds in an auction held on November 29th, 2011:
In this first stage of the auction, all dealers are required to submit both bids and offers on the bonds. They may have the bid or offer traded on, but they won’t know until the second stage.
One thing that can be immediately seen from the above is that without the auction, banks would indeed settle CDS at different prices, though the differences would have been slight in this instance — some auctions have better agreement in price than others.
The next step, still part of stage one, is to get an overall consensus price from the above submissions. This consensus price will be carried into the second stage of the auction.
To get the consensus, first one sorts the submissions to reveal the best bids and offers from a client’s perspective — in other words, sell high and buy cheaply from the dealer (or the highest bids and the lowest offers).
Then the best half of the marks are averaged to get the “Initial Market Midpoint”. For Dynegy, averaging all the marks in purple gives 69.5. (Note that all of the bids and offers in the chart below are the exact same as in the first chart; they have merely been shifted around to reflect the highest bids and lowest offers at the top.)
Crossing marks — offers lower than the bid — are not included in the calculation, hence the cross out at the top.
By now, you may be asking what keeps the banks honest in the above. Well, there is penalty for being off-market, as measured by the IMM (here 69.5) and by the direction of the net demand to buy or sell bonds. We’ll explain the penalty and how it’s calculated in a moment.
Stage one of the auction includes another, separate process independent of finding the Initial Market Midpoint.
This other process involves dealers submitting their requests to buy or sell bonds in order to establish demand. Remember that dealers may be acting on their own account here, or on behalf of their clients. For example, maybe Barclays doesn’t want to buy bonds but a client of theirs does, so the bank acts as a conduit for the client’s trade and submits a bid.
Here are the requests that were submitted for Dynegy bonds:
Note that the title “Physical Settlement Requests” is misleading. There is no physical settlement of CDS going on here. This is just about buying and selling bonds. The only connection to physical CDS settlement is that one could choose to replicate the transfers of physical settlement by trading the bonds in the auction and cash settling. Like so:
In the case of Dynegy, it can be seen that banks want to buy bonds more than they want to sell — that’s what it means for the net open interest to be $61.156m to buy.
Now we’ll explain why this matters for how the dealers get penalised if they submitted bids and offers that diverge too much from the consensus price (the Initial Market Midpoint). The simplest way to think about this is that if a dealer is on the wrong side of the consensus price where there isn’t demand, there is a penalty to pay.
Here’s what we mean. Because there is a net open interest to buy in the Dynegy example, you can logically expect the price of the bonds to climb above the consensus price in order to satiate the excess demand.
Let’s look at Citi’s prices and its request to trade bonds. They were willing to buy bonds at more than the consensus price (70 69.5). That makes sense because, as we just explained, the net open interest is to buy. They were on the right side of the consensus price given that the net demand is to buy, not sell.
On the other hand, the submission by RBS makes far less sense. Their offer suggests they were willing to sell at 31.6 even though the net demand was to buy, not sell. In the financial world, the submission from RBS is known as a “cock up”, and it results in a fine. The fine is equal to the difference between the consensus price and the RBS offer price (0.695 – 0.316) multiplied by the standard size for all trades submitted in the first round ($5m).
$5m * (0.695 – 0.316) = $1.9m, and the fine is known as an “adjustment amount” in polite company (but we don’t mind if you stick with “cock-up”).
By the way, sometimes such adjustments aren’t the result of cock ups. Sometimes they are imposed simply because the bonds rather illiquid, making it hard to know where to price them. When this is the case, the amounts are smaller, more normal — as in the Irish Life and Permanent auction in July:
Biscuits!!
Once the consensus price and the list of shame is published, it’s time for a few hours’ break for tea and cookies.
Stage Two
Once everyone is back, it’s time to start filling the $61.156m of net open interest to buy. But let’s look at how that net open interest arose just one more time (same data as the above “Physical Settlement Requests” just with different formatting):
If the banks tried to trade bonds with each other now, there would be some rather disappointed buyers. Stage two is all about trying to find sellers for the $61.156m so that those buyers aren’t sad that they didn’t get what they wanted.
There are two ways that those sellers are going to be found for the buyers.
First, all of the offers are going to be carried forward from stage one to stage two, at the standard trade size of $5m. Participating dealers are required to have at least $5m of bonds on their books to sell. The fact that they may be required to trade on their submitted prices from stage one is another mechanism that keeps everyone honest.
Second, the dealers can enter the sizes and prices at which they are willing to do to some additional selling above and beyond the trades sourced from their stage one prices. These additional trades are known as “limit orders”.
After all the data is entered — offers from stage one and limit orders from stage two — it’s time to rank them from the best to the worst prices and start filling the $61.156m of net open interest to buy.
If you had a cock-up in the first round, as RBS did, the Initial Market Midpoint will be considered the offer you meant to write — hence the replacement in the below done to preserve RBS’ modesty:
Here’s the same table, but with only the trades that ultimately went to determine the final price. Again, the goal is to find an additional $61.156m worth of bonds that can be sold to the eager buyers:
Also listed in the above table is which part of the auction the offer and size came from, though the asterisks also tell you this. The final stage two limit order, from Deutsche Bank, is only partly filled, hence the little carrot ^ by it.
Now that enough additional sellers have been found, everyone can trade bonds with each other, and they all do so at the final price of 71.25, as that is considered the equilibrium at which enough additional sellers were found.
Dynegy bonds ultimately settling at 71.25 in the auction meant that for every $10m of CDS notional, a protection buyer would have received $2.875m.
One final anti-abuse mechanism that has been built into auctions is the concept of a cap. In the case of Dynegy, let’s say the price of the last limit order used to fill the net open interest had been one percentage point below the consensus price, or 68.5. Then 68.5 would have been taken as the final price. This type of limit was put in place after some dodgy results in the Thomson auction in 2009. Whoops.
__________________________
To summarise, though, let’s have a reminder from trade body Isda about the main goal of auctions:
Prevents price and settlement distortions that could result where CDS exposure to a particular entity is disproportionately larger than the amount of actual obligations of such entity.
In other words, preventing short squeezes like Delphi. We’ll leave you to the philosophical debate on whether having CDS out there that are multiples of an entity’s debt outstanding is a good or bad thing.
As previously mentioned, however, further technicals around the deliverables have plagued auctions. Spot what’s wrong with this settlement:
Why are the sub bonds worth more than the senior ones? (The different tiers of debt have different auctions.) The problem is that if there aren’t enough bonds to trade in the auction, it might produce recoveries that are too high. In that sense, even auctions aren’t completely immune to their own short squeezes.
And as Satyajit Das pointed out in his post back in 2009 it was also interesting that the two entities settled differently:
Differences in the payouts between the two entities are also puzzling given that they are both under identical “conservatorship” arrangements and the ultimate risk in both cases is the US government.
Well, at least it inspired some CDS merc:
We’re copyrighting that one…
Related links:
Thomson CDS holders may boycott ISDA auction – FT (2009)
Credit Derivatives Determinations Committee website – Isda
Hammer it home – Markit
ISDA Credit Default Swap Auction Protocols – DechertOnPoint