Goldman Sachs Abacus 2007-AC1

A hedge fund wants a bespoke short

In mid-2006, the hedge fund Paulson & Co. approached Goldman Sachs wanting a large short position on the US subprime mortgage market. Paulson & Co. believed that a wave of subprime mortgage bonds, sold to investors during the housing boom, was heading for default. Shorting the bonds directly is hard to do at scale, so the fund wanted something else: a bespoke deal built specifically so that betting against it would be straightforward.

Goldman built that deal and named it Abacus 2007-AC1, a synthetic collateralised debt obligation (CDO) referencing a portfolio of about 90 subprime residential mortgage-backed securities (RMBS), with a notional value of about US$2 billion. Between January and March 2007, Goldman sent ACA Management, a firm with a reputation for picking bond portfolios, a list of roughly 123 candidate RMBS. Paulson & Co. and ACA then discussed and negotiated the final list together, with ACA keeping around half of Goldman's suggestions and adding others of its own.

A conflict left undisclosed

ACA was not told, and neither was IKB Deutsche Industriebank, the German bank that would end up holding most of the long side, that Paulson & Co. was on the other side of the trade. Paulson & Co. was not a neutral adviser helping to build a sound portfolio; it was the party that had approached Goldman wanting to short exactly this kind of asset, and it had every incentive to help select bonds likely to fail. Goldman's own marketing materials described the portfolio as having been "selected by" ACA Management, which was true as far as it went, but the materials omitted Paulson & Co.'s role and its opposing economic interest entirely.

Close, then collapse

The deal closed on 26 April 2007. IKB invested through a Goldman-arranged note structure; ACA Capital, ACA Management's parent, took on the CDS exposure on the long side; Goldman earned about US$15 million for structuring and marketing the deal. Paulson & Co. sat on the other side of the same CDS book as the protection buyer, paying premiums and waiting.

The collapse came quickly. By around 24 October 2007, roughly six months after closing, about 83% of the referenced mortgage bonds had been downgraded. By January 2008, about 99% had been downgraded, and the deal that ACA and IKB had bought into was, for practical purposes, worthless. Long investors lost over US$1 billion between them: ACA's loss is put at about US$900 million and IKB's at about US$150 million. Paulson & Co.'s short position earned it about US$1 billion, a gain almost exactly the size of the combined losses on the other side, the symmetry expected of a trade that is close to zero-sum once dealer fees are set aside.

The fraud case

On 16 April 2010, the US Securities and Exchange Commission (SEC) filed a civil fraud complaint against Goldman Sachs & Co. and Fabrice Tourre, the Goldman vice-president who had structured and marketed the deal. The complaint was narrow in scope: it was not that shorting subprime bonds was wrong, and not that ACA had done a poor job picking bonds. The charge was that Goldman's marketing materials contained a material omission, telling investors the portfolio was selected by an independent manager without disclosing that a party betting against the deal had shaped that selection.

On 15 July 2010, Goldman settled with the SEC for US$550 million, at the time the largest civil penalty the SEC had obtained from a Wall Street firm. Of that, US$300 million went to the US Treasury and US$250 million was returned to harmed investors through a Fair Fund. Goldman did not admit or deny wrongdoing, but it did acknowledge that its marketing materials "contained incomplete information."

Tourre's case went further than Goldman's. The SEC's claims against other individuals involved were dropped, but Tourre went to trial, and on 1 August 2013 a federal jury in Manhattan found him liable on six of seven civil securities fraud charges. On 12 March 2014, the presiding judge ordered him to pay about US$825,000 in total, a civil penalty of about US$650,000 plus disgorgement of about US$175,000 linked to a bonus he had received for the deal, below the roughly US$1.15 million the SEC had sought.

The CDS underneath the deal

Abacus sits directly on Lecture 7's material on credit default swaps (CDS), because that is exactly what this deal was built from. Abacus was not a cash CDO holding real mortgage bonds; it was a synthetic CDO, meaning the whole reference portfolio existed only as a set of CDS contracts written against those 90 mortgage bonds. In a CDS, the protection buyer pays a periodic premium and receives a payout if the reference asset defaults or is downgraded sharply, while the protection seller collects the premium and pays out if that happens. Selling protection is economically equivalent to being long the reference credit, in the same way an insurer is paid a premium for taking on a risk of claim.

Paulson & Co. was the protection buyer, meaning it was short the pool: it paid premiums and gained heavily when the underlying bonds defaulted. IKB and ACA were, in economic substance, protection sellers, meaning they were long: they collected premium income and bore the loss when those bonds failed. The tranching logic from earlier in the course applies directly here. A CDO's total expected loss is fixed by the quality of the loans in the pool, not by how the claims on that pool are sliced into tranches. Tranching decides who gets paid first and who absorbs losses first; it does not make the underlying pool any safer. Abacus took this one step further, by letting the party betting against the pool have a hand in choosing which loans went into the pool in the first place. That is what converts an ordinary model-risk question, whether the assumed default correlations were realistic, into an adverse-selection and disclosure problem: whoever chooses the pool has every reason to pick the worst available loans if that party is the one who profits when those loans fail.

The regulatory case itself stayed narrow, and that is exactly how the SEC argued it: the case turned entirely on disclosure. The SEC did not argue that a short bet on subprime housing was improper; shorting a security believed to be overpriced is a legitimate, everyday position. Nor did the SEC argue that ACA's picks were poor picks in themselves. The argument was that Goldman's marketing materials told IKB and ACA the portfolio was "selected by" ACA, without telling them that Paulson & Co., a party with the opposite economic interest, had helped shape that same selection. Once that omission sat in the marketing materials, a legal conflict of interest, a hedge fund shorting a deal it was not involved in structuring, turned into an alleged fraud, because the investors were misled about who was really steering the choice of assets they were being asked to trust.