Felix Salmon

Did ABN Amro hedge its ACA exposure?

By Felix Salmon
April 26, 2010

I just found this, buried in a Landon Thomas story on Abacus at the end of last week:

Some inside ABN Amro were leery of Abacus early on. Indeed, several traders there immediately bought credit default swaps — insurance-like instruments — from Goldman Sachs to hedge their exposure to ACA.

Now it’s worth noting that the headline on the story is “A Routine Deal Became an $840 Million Mistake”, and Thomas repeats many times in the story that the deal ended up costing Royal Bank of Scotland, which bought ABN Amro, $840.1 million.

But the bit about the CDS hedge is intriguing — especially the detail that the hedge was bought directly from Goldman Sachs, which was the company hedging its own exposure in the first place. If Goldman was happy to write credit protection on ACA, and presumably buy offsetting ACA protection elsewhere in the market, then why didn’t it just use ACA to insure the Abacus deal directly, and then turn around and hedge its ACA counterparty risk?

The answer, I think, might have something to do with collateral posting: Goldman only wanted CDS hedges where it could ask for collateral in the event of a downgrade or a fall in market prices, while ACA only wanted to insure the deal if it didn’t risk needing to post hundreds of millions of dollars with Goldman. So Goldman went through ABN Amro instead, which was happy to agree to Goldman’s collateral requirements.

I’m not clear how the $840.1 million was calculated — was that net of the hedges, or was that before Goldman paid out on them? And if ABN was hedging its ACA exposure with Goldman, why did it make any sense to do the deal in the first place? I’m sure that Goldman charged ABN Amro more than 17bp for ACA protection, so it’s hard to see how a hedged ABN could make any money on the deal at all.

Maybe if and when the UK government sues Goldman Sachs for RBS’s losses, we’ll learn more. But right now it’s all very vague where those super-senior losses really did ultimately end up.

13 comments so far | RSS Comments RSS

The ABN-GS CDS hedge on ACA is briefly discussed in the first of the two Goldman defense letters to the SEC. Here’s the relevant bits (from page 29 of the letter):

“ABN purchased protection from Goldman Sachs in the form of a $27 million corporate CDS referencing ACA’s credit rating. To Goldman Sachs’ knowledge, ABN did not purchase credit protection referencing the Reference Portfolio.”

Couple things worth pointing out:
1. The $909M supersenior (50-100) trade intermediated by ABN for 17bps was effectively a dual-default swap (or a contingent CDS). That is, ABN would only be on the hook if (and to the extent that) BOTH of the following credit events occurred: (a) the 50-100 tranche of the Abacus reference portfolio experienced a loss and (b) ACA defaulted.
2. ABN did not perfectly hedge this with Goldman. Instead, they entered into what appears to be a vanilla CDS on ACA with a notional of $27M.

This leads me to believe that ABN was attempting to delta-hedge the dual-default ACA/Abacus credit risk, leaving them exposed to correlation and jump-to-default risk. How much ABN ultimately lost on this trade, net of hedging, depends on the success of such a delta hedging strategy.

Posted by Sandrew | Report as abusive

“And if ABN was hedging its ACA exposure with Goldman, why did it make any sense to do the deal in the first place?”

This question assumes that the interests of the persons who decided to do the deal are aligned with those of the corporation as a whole. Surely the last couple of years (or decades…) have shown that is not always the case. It’s possible that the employee(s) profited handsomely through commissions and bonuses, even as they exposed their corporation to an unnecessary risk.

Posted by KenInIL | Report as abusive

How does a 27m CDS protect ABN from 1 yard (notional) of risk? I guess the dual default made it seem less risky, rather than more risky?
How does 17bps on a 909M SS trade for (I assume) 3.5 WAL come back to 7mio profit as mentioned by NYT? I assume it was upfront fee, but was the maturity seen as longer than the WAL of the SS?

Posted by Dukey | Report as abusive

Another trip down the GS cayman rabbit hole in pursuit of balance-sheet logic? You can speculate all day where those elusive supersenior losses might actually have arisen but chances are you’d have better luck finding WMDs in Baghdad.

But you’re right that it’ll get clearer when Europe does the heavy lifting of a real prosecution. By that I mean a sophisticated prosecution: criminal and civil, with a twist of human rights violations thrown in. I wouldn’t speculate on this not happening.

Posted by HBC | Report as abusive

Only CDS that ABN could have bought to hedge the structure would be a CDS on ACA. Buying a CDS on another CDO would just mean paying for basis risk.

And it would be a nifty if GS did sell a CDS on ACA to ABN which GS in turn purchased from esteemed, revered client.

Posted by longandshort | Report as abusive

Please forgive my non-financial naïveté, but if anyone could de-confuse me on the following I’d be grateful:

1. Who OWNED the underlying portfolio of assets that Paulson/ACA selected and which subsequently defaulted? If they were chosen from GS’s books, mustn’t GS have taken a massive hit too (equal to that of ACA)when they tanked? Wouldn’t they have done better keeping the Abacus credit protection from ACA for themselves, rather than selling it in to Paulson? (Perhaps they hedged these assets with other investors outside of the Abacus structure?)

2. If they didn’t own these assets, where/how did Paulson find them?

3. Still if they didn’t own the assets, did ACA and IKB know that the CDSs making up Abacus were thus “naked” from the start? Perhaps they thought they were “covered” CDSs, i.e. that they weren’t entering into a straight bet with a short investor, but simply providing credit insurance to an investor who had no intrinsic interest in seeing the underlying assets default?

Thanks for the help!

Posted by hillemarc | Report as abusive

“If Goldman was happy to write credit protection on ACA…”
they weren’t and didn’t, they wrote protection on ACA *credit rating* (not my definition of a plain vanilla CDS btw, as Sandrew wrote above)

“I’m sure that Goldman charged ABN Amro more than 17bp for ACA protection”
when you buy insurance or in this case credit enhancement, you don’t charge, you pay, so GS paid ABN AMRO 17bps to enhance ACA’s credit.

Which all goes to show that ABN AMRO, was hardly negative on ACA as they “hedged” only for a credit downgrade and not for a normal credit event like failure to pay or bankruptcy and didn’t bother to hedge the reference porfolio at all.

Posted by alea | Report as abusive

ABN AMRO is going to end up looking foolish here, as will the SEC

Posted by STORYBURNcom_0 | Report as abusive

“they wrote protection on ACA *credit rating*”

Huh. What is that called, a rating migration swap? That’s bizarre. Ignore my previous comment; I have no idea what ABN was thinking.

Posted by Sandrew | Report as abusive

Unusually, I don’t think I can agree with Sandrew here, unless I have completely the wrong end of the stick.

First, this buying 27mm protection on ACA is a “sun rises in the east” story. A modern iBank dynamically hedges all of its OTC counterparty exposure as measured by bilateral credit valuation adjustement (CVA). The way this works is that there is a central CVA desk that essentially sells credit insurance to all of the other trading desks in every asset category: FX, IR, equity, energy, credit, whatever. There is a kind of internal transfer pricing scheme for this insurance that may be reflected in the prices made by the business desks. So what probably happened is that ABN crunched the numbers and found that the GS/ACA deal moved their ACA CVA by 27mm. In that case, that they bought protection from GS would have been a coincidence and completely different traders involved from the Abacus deals.

Note that CVA hedging is a dynamic process; it must be continually updated not only as new positions arrive but as the MtM of current positions changes. CVA traders usually try to hedge some of their MtM sensitivities to other asset classes, but that would not have been practical here because of the illiquid underlying. The way you get nailed for 800+mm of loss when you are “hedged” is if the MtM changes quickly in tandem with the price of credit protection.

Why the weirdo “credit rating” reference terms? I dunno – did ACA have any bond issues that could have served as reference securities?

Finally, what is described by Felix was not a contingent CDS (sorry, I haven’t had time to look at any of the primary docs.) The “contingent” part means that the amount of protection is contingent on the MtM of some reference portfolio (all before recovery rates are applied.) But here the 27mm notional was fixed. If it *had* been a contingent CDS sold by GS to ABN – that really wouldn’t have made sense!

Posted by Greycap | Report as abusive

Just realized: I kept writing “MtM” when I meant “PFE”.

Posted by Greycap | Report as abusive

Greycap, maybe I have this wrong, but I don’t think our interpretations are that far off.

I think it’s apparent from the record that ABN’s $909M notional supersenior position amounted to a contingent CDS. The contingency part (the reference swap) is the MtM of the $909M supersenior (50-100) tranche of the $1.8B Abacus portfolio. The CDS part (the reference obligor) is ACA default risk. Whether it was structured as a CCDS we don’t yet know, but that’s moot. For example, it might simply be the case that ABN is literally an intermediary between Goldman and ACA on the supersenior tranche trades, where with the former ABN agreed to ACA’s no-collateral terms and with the latter ABN agreed to Goldman’s demand for (bilateral?) collateral posting. Please correct me if you see it differently.

We’re speculating as to the why’s and how’s of ABN’s hedge of this position. I agree that, whatever the form, ABN’s net supersenior position is functionally equivalent to retention of ACA counterparty risk. So it would make a lot of sense that ABN’s CVA desk would be tasked with determining how much of a hedge to put on to cover their mean negative exposure to ACA on the supersenior trade. Good observation on your part.

Another way to think about this trade is to go back to the motivations and roles of the parties. Goldman was basically acting as a broker between Paulson and ACA. Goldman had a limited appetite to retain any jump risk, so they demanded that ACA agree to pony-up collateral if/when the MtM on the supersenior moved against them. ACA said no. ABN decided it had the appetite for the jump risk and agreed to intermediate. 17bps is the price ABN charged for retaining ACA counterparty risk. If they hedged this with a CCDS (whether to Goldman or anyone else), then ABN would have been acting as a broker of counterparty risk, and that wouldn’t really make sense. Rather, ABN is acting as a market-maker. They take on the ACA counterparty risk freely, and dynamically hedge it (likely out of the CVA desk) as they do other bits of counterparty risk. Given the speed with which the subprime market tanked, the illiquidity of the underlying Abacus tranche, and the wrong-way risk (correlation b/w ACA and Abacus tranche), it’d be no surprise if ABN’s hedge failed.

Of course, I’m at a loss to explain the bizarro rating migration term, which makes me suspicious that I’m possibly way off base on this whole mess.

@dukey: Take what I say with a dose of salt. There’s a good chance I’m an idiot.

Posted by Sandrew | Report as abusive

Apologies, Sandrew, I was guilty of some careless reading and thought you were saying the protection ABN bought amounted to a CCDS. Agreed, the protection they wrote to Goldman was equivalent to a CCDS.

I am also guilty of some careless writing: I should make clear that a CVA hedge is a hedge of expectations, not risk. The fact that any given hedge “fails” is therefore not evidence in itself that it was wrong. Nonetheless, I would be surprised if it were right in this case because the incompleteness and illiquidity of the reference markets makes it pretty hard to come up with a unique and verifiable expectation in a pricing measure.

Posted by Greycap | Report as abusive

Post Your Comment

We welcome comments that advance the story through relevant opinion, anecdotes, links and data. If you see a comment that you believe is irrelevant or inappropriate, you can flag it to our editors by using the report abuse links. Views expressed in the comments do not represent those of Reuters. For more information on our comment policy, see http://blogs.reuters.com/fulldisclosure/2010/09/27/toward-a-more-thoughtful-conversation-on-stories/