All The Devils Are Here: Unmasking the Men Who Bankrupted the World (52 page)

Goldman’s chief risk officer, Craig Broderick, would later say that “our client base is extremely aware and clear about what function we are performing.” But contemporaneous e-mails—and complaints after the fact—would paint a messier picture.

The structured product group, which was part of the mortgage desk and which put together many of Goldman’s most complex trades, was co-headed by Michael Swenson, a former Williams College hockey player in his late thirties, and David Lehman, a young star hired from Deutsche Bank around 2004, when he was just thirty. They made a decision early on that synthetic business could be big, so they staffed up accordingly. Or, as Swenson put it in his 2007 self-evaluation, “I can take credit for recognizing the enormous opportunity for the ABS synthetics business two years ago. I recognized the need to assemble an outstanding team of traders and was able to lead that group to build a number one franchise.” And that he did. Among those he recruited was Josh Birnbaum, a star trader who traded the new ABX index.

Until synthetics came along, Goldman had been a middling player in both the mortgage-backed securities and the CDO markets. But it quickly became a force in this new market, which consisted of both synthetic CDOs and hybrid CDOs. (These contain both real mortgage-backed securities and credit default swaps.) Quickly, Goldman began to climb up the rankings; in both 2006 and 2007, it was the fourth largest underwriter of CDOs. Wrote Birnbaum in his 2007 self-review: “#1 market share in ABS CDS [ABX and single name] est 30–40%.” Although mortgages were a relatively small piece of Goldman’s overall business—the department’s 2006 revenue barely topped $1 billion, compared to nearly $38 billion for the firm itself—almost half of the mortgage desk’s 2006 revenue came from “structured products trading and CDOs.”

One of the earliest Goldman synthetic CDOs, put together in 2004, was called Abacus. It came about when IKB, a German bank that had become an aggressive CDO investor, came to Goldman seeking exposure to a specific set of mortgage-backed securities. Goldman was happy to oblige, and built a synthetic CDO based on the securities IKB wanted to reference. The dollars IKB received would come from the customers who took the short side
of the transaction. Over the next few years, Abacus became a kind of Goldman franchise, with about sixteen deals and $10 billion worth of securities sold, according to the Senate Permanent Subcommittee on Investigations.

Later, Goldman would insist that the synthetic CDO deals the firm put together were “often initiated by clients,” to quote Goldman’s 2010 letter to shareholders. In that original Abacus CDO, this was clearly true. “IKB craved this product,” says a person familiar with the deal. He adds that, in the early years, the hard part was finding someone to take the short side of the trade, because no one wanted that risk.

But the firm’s later insistence that it was merely a “market maker” in these transactions—implying that it had no stake in the economic performance of the securities it was selling to clients—became less true over time. And even those early deals made a mockery of the notion that most investors understood what they were buying. For one thing, the synthetic deals were stuffed full of multiple kinds of risks. Take Abacus 2005-3, another early deal. The reference obligations—that is, the securities the CDO referenced—consisted of 130 credits. Those credits included everything from tranches of mortgage-backed securities issued by Long Beach Mortgage, Countrywide, Ameriquest, and New Century, to commercial mortgage-backed securities, to trusts backed by Sallie Mae student loans, to credit card debt issued by MBNA and Chase. Who could possibly understand all the risks contained in these securities? Is it any wonder that investors were essentially buying ratings instead?

Nor was it possible for investors to know Goldman’s own position. Was Goldman merely standing between two customers, the way a true market maker did? Or was Goldman using—and designing—the CDO in order to hedge an existing position or to “express” its own view about the referenced securities? For instance, Goldman could attempt to profit by having, say, a long position in triple-A-rated securities, and then use a synthetic CDO to establish a short position in triple-B-rated securities. Or, if Goldman was long in New Century mortgage-backed securities, the firm could hedge its position by constructing a synthetic CDO that referenced those securities, and take the short side of that trade. Buyers had no way of knowing whether Goldman was the dealer at the card table, a player, or both. Goldman supporters would later argue that buyers shouldn’t have cared what Goldman’s position was—they were responsible for doing their own analysis of the underlying securities. “The deal is the deal,” Dan Sparks, the former head of the Goldman mortgage desk, later told the Senate Permanent Subcommittee.
But many buyers didn’t do that analysis. And it would later become clear that at least some certainly
did
care what Goldman itself was doing.

Just as a cash CDO was the perfect mechanism for laundering risky mortgages, so was a synthetic CDO the perfect vehicle for laundering positions a firm wanted to get off its books. For instance, if you owned a bunch of New Century mortgage-backed securities, it might be hard to find someone who wanted to own the risk that New Century loans were going to go bad. But if you could establish a short position for yourself by selling your clients the long position in a synthetic CDO—thereby insulating yourself from your unwanted junk by creating new triple-A securities—well,
voilà
! Given that Goldman was a big warehouse lender to New Century—and far more likely than its clients to have early knowledge that New Century mortgages were doomed—the whole edifice begins to take on a very dark hue. As Janet Tavakoli, a structured finance expert who became a fierce, prescient critic of CDOs, later put it, “They had reason to know what they were hedging shouldn’t have been created in the first place.”

In addition, Goldman, says one person who has looked into these deals, “had a stranglehold on every aspect of the transaction.” The fine print of one Abacus prospectus says that the “Protection Buyer”—i.e., Goldman—“may have information, including material, non-public information,” which it did not provide to the buyers. Thus, if Goldman did have knowledge of New Century mortgage defaults, it was under no legal obligation to share that knowledge with a client who was about to buy a synthetic CDO that referenced New Century securities. In some of the Abacus deals, Goldman could unwind the trade after three years if it didn’t like how it was going. (One blogger later called this “Heads I win, tails you lose.”) Another prospectus notes that Goldman’s many roles “may be in conflict with the interests of the investors in the transaction.” In fairness to Goldman, these pitfalls were identified in writing. But the disclosure didn’t make the conflicts go away.

The point is, investors were hardly buying an existing security from a neutral market maker. They were buying a security that had been constructed to enable someone to accomplish a specific goal. Quite often, that “someone” was the market maker itself. As Goldman’s Tourre wrote in a June 2006 e-mail, “ABACUS enables us to create a levered short in significant size.” (It was levered because Goldman didn’t have to put up cash.)

And there were times when it appears Goldman wasn’t a market maker at all, but rather a principal. Consider this exchange between David Lehman and Josh Birnbaum: “[We] need to decide if we want to do 1–3bb of these
trades for our book or engage customers,” wrote Lehman. Replied Birnbaum, “On baa3 [the lowest rung of investment grade], I’d say we definitely keep for ourselves. On baa2 [the second lowest rung], I’m open to some sharing to the extent that it keeps these customers engaged with us.”

It is also clear from internal e-mails that by 2006 there were Goldman traders—not all of them, but some—who viewed some of their subprime holdings as junk. One trader described a Goldman mortgage-backed security this way: “It stinks…. I don’t want it in our book.” Swenson later wrote in a self-review that “during the early summer of 2006, it was clear that the market fundamentals in subprime and the highly levered nature of CDOs was going to have a very unhappy ending.”

The year 2006 was when Dan Sparks became the head of the mortgage desk. Sparks, an intense Texan who had joined Goldman as an analyst in 1989 after graduating from Texas A&M, spent his early years at the firm helping the Resolution Trust Corporation dispose of assets from the S&L crisis. He made partner in 2002. Sparks was a classic trader: he didn’t let a lot of hope, fear, or sympathy creep into the equation. The price was what the market said it was, and if you were willing to buy securities at that price, whatever happened after the sale was your responsibility. It wasn’t Goldman’s job to protect clients from their own mistakes, he believed; they were all big boys. Whatever the firm’s purpose was in constructing a particular synthetic CDO—market maker, principal, whatever—was irrelevant. In protecting Goldman’s interests, that would prove to be a useful attitude.

Consider, for instance, a November 2006 deal called Hudson Mezzanine, a synthetic CDO that referenced triple-B subprime securities. In terms of serving Goldman’s interests—in a way that wouldn’t be obvious to investors—it was a classic.

The CDO had been constructed, Goldman executives later told the Senate Permanent Subcommittee, while the company was trying to remove triple-B assets from its books. Among those assets was a long position in the ABX index that Goldman had gotten “stuck” with while putting together deals for hedge fund clients that wanted to go short. Unable to find counterparties to take the long position off its hands, Goldman used Hudson as a means by which it hedged its long position.

Goldman selected all the securities that Hudson would reference. These included $1.2 billion in ABX index contracts, offsetting the long ABX position Goldman wanted to hedge, and another $800 million in single-name
CDS, or credit default swaps that referenced specific mortgage-backed securities issued in 2005 and 2006.

None of which was clear from the Hudson prospectus. Instead, the disclosure merely said that the CDO’s contents were “assets sourced from the Street,” making it sound as though Goldman randomly selected the securities, instead of specifically creating a hedge for its own book. Page four of the pitch book also said, “Goldman Sachs has aligned incentives with the Hudson program by investing in a portion of the equity.” Only on page thirteen does the pitch book make the standard disclosure that Goldman was providing the initial short position. But according to the Senate Permanent Subcommittee, Goldman didn’t then sell that short position to a client, as a true market maker would. Instead, “Goldman was the sole buyer of protection on the entire $2 billion of assets,” as an internal Goldman e-mail put it.

If the job of a market maker is to sit between two investors, each of which affirmatively wants to take a different view, the Hudson deal didn’t come close to that definition. Nor is it possible to argue that Goldman was doing something its clients were clamoring for. Rather, it was a deal Goldman had to sell, and sell hard, to reluctant clients. Swenson would later write about such deals, “[W]e aggressively capitalized on the franchise to enter into efficient shorts….”

Later, as Hudson and other deals went sour, Goldman’s clients were furious at how they had been taken in. “Real bad feeling across European sales about some of the trades we did with clients,” wrote Yusuf Aliredha, Goldman’s head of European fixed-income sales, in an e-mail to Sparks. “The damage this has done to our franchise is very significant.”

On December 5, 2006, just before noon, Dan Sparks sent an e-mail to his bosses, Tom Montag, head of sales and trading in the Americas, and Rich Ruzika, co-head of global equity trading. “Subprime market getting hit hard—hedge funds hitting street, wall street journal article.” (He was referring to a
Wall Street Journal
story published that morning about how subprime borrowers were increasingly falling behind on their mortgages.) “At this point we are down $20 mm today. Structured exits are the way to reduce risk.”

More and more traders on the mortgage desk were getting increasingly uncomfortable being on the long side of the mortgage markets—the Hudson
deal had shown that. Individual traders were even shorting mortgage securities. Yet as a firm, Goldman still had a large overall long position. At the end of November, the firm had $7.8 billion in subprime mortgages on its balance sheet, and another $7.2 billion in subprime mortgage-backed securities. Goldman also had a big long position in the ABX index, as well as warehouse lines extended to New Century, among others. The gossip among traders at other firms was that Goldman Sachs was heavily exposed to the mortgage market—and they were right. But they couldn’t see inside Goldman Sachs.

Roughly a week after Sparks’s e-mail, CFO David Viniar convened a meeting in his office. Even though Goldman’s internal risk measures, such as VaR, suggested that everything was okay, the mortgage desk had nonetheless suffered losses ten days running. Gary Cohn, as president, would later testify that the firm’s internal risk models had “decoupled” from the actual results in December 2006. Goldman’s top executives were sensitive to the losses because the firm was fanatical about using mark-to-market accounting, valuing the securities it held daily, based on the price they would get if they sold the securities in the market that day. They reflected any gains or losses on the firm’s books, and reported this information to Viniar and Goldman’s other top executives. Unlike every other firm on Wall Street, in other words, Goldman had no illusions about how its mortgage-related securities were performing.