From the Horse's Mouth: AIG's Mistake Explained

Here's an interesting exchange with an AIG executive from the FCIC interviews. As I explained a while ago, what really killed AIG were the terms of its CDS contracts on subprime CDOs and RMBS — specifically, the contracts included Credit Support Annexes (CSAs) which required AIG to post collateral based on the mark-to-market value of the underlying CDOs. When the CDO shut down in 2007, mark-to-market prices collapsed as well, forcing AIG to post tens of billions in collateral on their CDS contracts.



In the FCIC's interview with Andrew Forster (mp3), who headed AIGFP's London branch, the interviewers — to their credit — actually directly asked asked Forster about this. His response is fascinating. The two FCIC interviewers were Chris Seefer and Dixie Noonan, and I've transcribed the exchange to the best of my ability, since it's pretty short: (emphasis mine)

Chris Seefer (FCIC): Did you know that the contracts for the multisector [CDS] book included terms that would require AIGFP to post collateral if there was a decline in the market value of the reference collateral?



Andrew Forster (AIG): Um, as I've looked back over my emails and notes, it's clear that there were times that I was copied on an email that would have that sort of information. Did I recall that? Did I look at it? No, I didn't. But it would be disingenuous for me to say that I clearly had no knowledge of it, because I must've, I should've had some back at that time. But it wouldn't really have rang any great — it wouldn't have resonated with me on the basis that my understanding was pretty much every derivative product that you write is gonna come with a CSA and a collateral agreement. That's just the market standard. So it wouldn't be odd if that was the case. But the exact specific terms of the different ones, no, I didn't really focus on it at the time.



...



Dixie Noonan (FCIC): You said that it was typical for the contracts to have collateral provisions. Was it typical, based on your understanding at the time, for the collateral provisions to be based on declines in the market value of the underlying securities?



Andrew Forster (AIG): Yeah I mean a typical derivative, or any credit derivative, they would all come — my understanding is that the market standard would be that that came with a CSA agreement and a collateral agreement, and that would be based on some estimate of market value for those products.



Dixie Noonan (FCIC): And would that be true not just for asset-backed security CDOs, but also for corporate CDS?



Andrew Forster (AIG): Yeah, I mean we wrote lots of other single-name corporate CDS. If you write one of those, it all comes with a collateral — I mean, 99% of the time it's gonna come with a collateral agreement. And it's based on the, you know, replacement value, or something like that, for the product.

OK, to be clear, it's simply not true that the "market standard" was for CDS on asset-backed securities to include full-blown CSAs, which require counterparties to post collateral based on mark-to-market prices. The only other big protection sellers for CDS on asset-backed securities at this time were the monoline insurers, which did not post collateral based on market prices. (The monolines did, however, agree to post collateral when their credit ratings were downgraded, but that's another story.) Whether protection sellers would have to post collateral based on market prices was a fairly contentious and prominent issue, and yet Forster was apparently unaware that a debate even existed. To him, CDS on asset-backed securities were just another derivative, and since most derivatives included full CSAs, CDS on asset-backed securities should naturally include CSAs too. (It sometimes took weeks to negotiate CSAs for bespoke derivatives, so I don't know why Forster thought all these deals should just include routine CSAs.)



Of course, there's a reason why the collateral-posting issue for CDS on asset-backed securities was so contentious: because a full CSA exposes the counterparties to significant market risk. So if you're a protection seller on a CDS, you would specifically want to avoid a full CSA when the underlying security trades in a particularly illiquid market, where prices jump around a lot, because that could require you to post large amounts of collateral on very short notice. AIG learned that the hard way.



That Forster gave this really very fundamental issue so little thought is just amazing.