Read the “synthetic, synthetic CDO” post first if you haven’t already.

The reasonable counterargument, for example here, is that because these are derivatives, there logically speaking must have been someone on the other side of the trade from the buyers, and the buyers should have known that — who that is doesn’t need to be disclosed. I think this is true to a degree, but not to the degree that Goldman needs it to be true.

Take an ordinary synthetic CDO. Back in 2005-2006, a bank might create one of these because it knows there is demand on the buy side for higher-yielding (than Treasuries) AAA assets. To do this, the CDO has to sell CDS protection on its reference portfolio to someone. That someone could in the first instance be the bank. But then the bank’s “short” position goes into its huge portfolio of CDS, which may overall be long or short the class of securities (say, subprime mortgage-backed securities) involved.The bank is constantly hedging that portfolio via individual transactions with other clients or other dealers, so there’s no one-to-one correspondence between the long side of the new CDO and any specific party or parties on the short side.

Let’s say for the sake of argument that the bank, prior to the new CDO, was exactly neutral on this market. The new CDO makes it a little bit short. So the bank will go out and hedge its position by finding someone else to lay the short position onto. But first of all, there’s a good chance it will divide up the short position and hedge pieces of it with different people. Those people may be buying the short position not because they want the subprime market to collapse; they might be partially hedging their own long positions in that market. Second, there’s an even better chance that it won’t sell off exactly the short position it just picked up from the CDO; it will buy CDS protection on a bunch of RMBS that are similar to the ones it just sold CDS protection on (which ones will depend on what the market is interested in), so in aggregate it comes out more or less the same.

So ultimately the “short” side of the CDO gets dispersed between the bank’s existing CDS portfolio and the broader market. So yes, there must be a short interest out there that is exactly equivalent to the long interest. But there doesn’t have to be a party or even an identifiable set of parties who have exactly the short side of the new CDO and want it to collapse, let alone a party that helped structure the CDO because it wanted to be on the short side. There’s a big difference between the market as a whole and one hedge fund.

Now, are things different with a synthetic synthetic CDO, as I have called it? Maybe. The pro-Goldman argument would be that ABACUS was so highly structured — basically, each tranche was a single complex derivative with a long side and a short side — that the long investors must have realized that there was a single party, or a small number of parties, on the other side. But that doesn’t necessarily hold. Just like a synthetic CDO, Goldman could have whipped this thing together because it thought it could sell it, and Goldman could have planned to hedge it the usual way — partially with its inventory and partially through a lot of small transactions dispersed throughout the market.

As always, I draw on Steve Randy Waldman.

By James Kwak

I actually suspected this, but I haven’t had the time to look at the marketing documents. But thankfully Steve Randy Waldman did. I don’t think I can improve on his description — these things take hundreds of words — but here’s a quick summary.

An ordinary CDO is a new entity that raises money by issuing bonds in tranches, uses the money to buy some other bonds (say, residential mortgage-backed securities) and uses the cash flows from those bonds to pay off its own bonds.

A synthetic CDO is similar except instead of buying the underlying bonds, it sells credit default swap protection on those bonds (the reference portfolio) and uses the premiums from the CDS to pay off its own bonds. (The money it raises by selling those bonds is usually parked in low-risk securities so it is available to pay off the CDS if necessary.)

ABACUS was different. There was a reference portfolio. But instead of selling CDS protection on all of those bonds, Goldman said (to paraphrase), “Imagine we sold CDS protection on all of those bonds. Then imagine we used those CDS premiums to issue bonds in tranches A-1, A-2, B, C, D, and FL. The derivative I’m selling you is one that will behave exactly as if it were an A-1  (or A-2) bond in that scenario — even though we’re not actually selling all of the tranches.”

Continue reading “ABACUS: A Synthetic, Synthetic CDO” →