All the Devils Are Here

by Bethany McLean; Joe Nocera

  It seems inconceivable now that O’Neal himself had so little understanding of what lay ahead. He was a very smart man, a tough, seasoned Wall Street executive. One of the formative experiences of his career had been the Long-Term Capital Management disaster. He had been Merrill’s CFO during that crisis, and it remained seared in his memory. He knew how a series of events could spiral into catastrophe. He remembered that awful feeling of realizing that Merrill couldn’t put a value on the collateral it held. He saw how, in a crisis, “everything is correlated”—meaning that securities that were supposed to act as hedges suddenly started falling in tandem, exacerbating the losses. Panics have their own momentum, their own rhythms. It didn’t matter how much cash you said you had; if your counterparties lost faith in you, you were finished. “You couldn’t rely on anything,” he liked to say.

  O’Neal had worked fifteen hours a day for months during the LTCM crisis. He had gone home, night after night, worried about whether the firm would have enough liquidity to fund itself the next day. And, he liked to say, he never forgot those lessons. Yet now it appeared as if he had forgotten those lessons.

  O’Neal had discovered another fact as a result of the Bear fiasco that should have shaken him to his core. He had learned the size of Merrill Lynch’s subprime exposure. It was enormous. When Kronthal had left in July 2006, the firm had somewhere between $5 billion and $8 billion in subprime risk on its books. Most of it was either subprime mortgages waiting to be securitized, tranches of mortgage-backed securities waiting to be put into CDOs, or triple-B CDO tranches waiting to be repackaged into new CDOs as triple-As. This was hardly an insignificant exposure; if those subprime securities had to be written down in large numbers, Merrill was going to feel a good deal of pain. People would undoubtedly get fired. But it was not an amount that could bring the firm down.

  A year later, Merrill Lynch held an astonishing $55 billion in subprime exposure on its balance sheet. In the space of one year, Semerci and Lattanzio had added somewhere between $45 billion and $50 billion in additional exposure. Some of it was the same kind of collateral that had been on the books when Kronthal had been running the show: mortgage-backed securities of one sort or another waiting to be resecuritized. But the vast majority of it was triple-A tranches of subprime CDOs.

  Anyone who looked closely at this triple-A exposure would realize in an instant what Lattanzio and Semerci had done. With AIG no longer around to write protection—leading to a lack of buyers for the super-senior tranches—the only way Merrill could continue churning out new CDOs was to keep the triple-A risk itself. So that’s what Semerci and Lattanzio had done. In the case of CDOs with subprime mortgage-backed securities, Merrill simply bought the triple-A tranches and put them on its books. In the case of synthetic CDOs—a business Merrill was also deep into—the firm would find a hedge fund to take the short position and take the long position itself. In most cases, Merrill bought protection from a monoline insurer like MBIA (which, under the rules, also enabled the firm to book the income on the triple-A tranches up front), but in the event of disaster, that wasn’t likely to help much. The monolines had insured so much triple-A risk that any market event that hurt Merrill Lynch would destroy them.

  The result of Semerci and Lattanzio’s strategy was that Merrill Lynch would remain the number one underwriter of CDOs and the two men would get their big bonuses. But in the process, they had put Merrill Lynch itself at grave risk.

  Did the two men understand that? At a certain point, late in the game, Semerci in particular seems to have understood the gravity of the situation. According to several former top Merrill executives, he appears to have managed his risk assumptions in such a way as to keep the estimated losses that he presented to management and the board artificially low. They also believed his marks were too high. These same executives are convinced, for instance, that Semerci knew full well when he made that board presentation in July 2007 that Merrill losses were going to be far higher than $83 million.

  But until it was far too late, it appears that Semerci and Lattanzio did not fully understand the import of their strategy. Why? Because just like Ralph Cioffi and Mike Tannin at Bear Stearns, Semerci and Lattanzio still believed that a triple-A rating meant something. As the market had gotten shaky, they had begun shorting the ABX triple-Bs, but it never occurred to them that they were on the wrong side of the triple-A bets. Their belief in the value of the triple-A was why Semerci could tell Kim, with a straight face, that Merrill had very little exposure to subprime risk: he still thought the triple-As were close to riskless. That’s why he could tell O’Neal he was reducing the risk in the portfolio. And that’s why Semerci and Lattanzio could estimate Merrill’s worst-case scenario losses in the tens of millions, rather than the billions. The two men were no different than the “real-money” investors who had been lured into the game by Wall Street, convinced they were getting the high-finance equivalent of a free lunch: an ultrasafe security that also generated higher yields than Treasury bonds. Even many of these investors, though, had become leery of a triple-A rating by the summer of 2007, especially as the spread between Treasuries and super-senior tranches narrowed to a smidgen. As Merrill Lynch had loaded up on triple-A mortgage-backed securities, the firm had become, without knowing it, one of the dumb guys. That was the real difference between Goldman Sachs and Merrill Lynch. “We fell for our own scam,” John Breit, the Merrill risk manager, would later say.

  For some time now, the synthetic CDO business resembled nothing so much as a daisy chain. It was just the way Lew Ranieri had described it in that speech he gave at the OTS. The buyers of the lowest-rated equity tranches weren’t investors who were eager to take that risk in return for the promise of a high yield. Many of those investors were gone. Mostly, the buyers were hedge funds interested in doing that correlation trade, the one where they bought the equity and then shorted the triple-A, so they won no matter what the housing market did. The riskiness of the equity slice was meaningless to them. The buyers of the mezzanine, or triple-B, slices were other CDOs, which would then launder them into new triple-A slices. And the buyers of the triple-A were quite often the underwriters themselves, taking the long side against the same hedge fund that had also taken the short side of the triple-As. With no need for actual collateral—since everything was referenced—such deals could be done ad infinitum. If you were working feverishly to churn out CDOs and keep your number one ranking, this was an important component of your strategy—because these were the easiest deals to do. So in addition to underwriting cash CDOs, using mortgage-backed securities, Semerci and Lattanzio also dove into the synthetic game.

  It was not a pretty thing to watch. Chicago-based hedge fund Magnetar would come to be the face of the correlation trade. According to the nonprofit investigative news service ProPublica, which conducted a six-month investigation into Magnetar’s trades, some $30 billion worth of CDOs in which Magnetar owned the equity were issued between mid-2006 and mid-2007; by J.P. Morgan’s estimate, Magnetar’s CDOs accounted for between 35 and 60 percent of the mezzanine CDOs that were issued in that period. Merrill did a number of these deals with Magnetar. The performance of these CDOs can be summed up in one word: horrible.

  The essence of the ProPublica allegation is that Magnetar, like Paulson, was betting that “its” CDOs would implode. Magnetar denies that this was its intent and claims that its strategy was based on a “mathematical statistical model.” The firm says it would have done well regardless of the direction of the market. It almost doesn’t matter. The triple-As did blow up. You didn’t have to be John Paulson, picking out the securities you were then going to short, to make a fortune in this trade. Given that the CDOs referenced poorly underwritten subprime mortgages, they had to blow up, almost by definition. That’s what subprime mortgages were poised to do in 2007.

  Take a deal called Norma, a $1.5 billion synthetic CDO that Merrill Lynch put together in March of 2007, and which would later be dissected by the Wall Street Journal.
The CDO manager Merrill chose to manage the deal was NIR. It was a former penny stock operator that Merrill had found and put into the CDO management business. Merrill had a number of similar captive CDO managers who knew without being told what kind of collateral the CDO was supposed to reference. Norma included a handful of subprime mortgage-backed securities—about $90 million worth, or 6 percent of the overall holdings, according to the Journal. It also included pieces of other CDOs, primarily triple-B mezzanine tranches, some of which Merrill had warehoused in order to launder them into new triple-A tranches at a later date, and some of which were being managed by CDO managers Merrill had hired—including Ricciardi himself, who had joined Cohen & Company, a big CDO manager. The rest of Norma consisted of credit default swaps that referenced tranches of other CDOs that contained subprime securities. In early 2007, all three rating agencies gave 75 percent of Norma’s tranches a triple-A rating.

  Magnetar bought the equity portion, of course. At the same time, it shorted the triple-A tranches of Norma, just as John Paulson had done in his Abacus deal. Merrill Lynch prepared a seventy-eight-page pitch book to help convince investors to buy pieces of the CDO. The Journal would later note that the pitch book stressed that mortgage-backed securities “have historically exhibited lower default rates, higher recovery upon default and better rating stability than comparably rated corporate bonds.” Merrill’s fee was in the neighborhood of $20 million.

  Ultimately, Merrill was able to sell $525 million worth of tranches, most of them lower-rated ones, which Merrill Lynch was promising at 5.5 percent interest above Libor, a very high yield. (Libor is the interest rate that banks charge when they lend to each other.) This was so even though, according to a lawsuit later filed against Merrill for its role in underwriting Norma, the securities had declined by 20 percent even before the deal closed. By December 2007—just nine months after Norma had been created—most of the deal had been downgraded to junk by the rating agencies.

  “It was a tangled hairball of risk,” Janet Tavakoli, the CDO critic, told the Journal. “In March of 2007, any savvy investors would have thrown this… in the trash bin.”

  But wait. If it was a $1.5 billion CDO, and Merrill could sell only $525 million of it, what happened to the other $975 million of Norma—all of which was triple-A? That’s what went onto Merrill’s books; it took the long position on the triple-As. This was the exposure that Semerci was claiming was nearly riskless.

  A lawsuit would later claim that Merrill was actively seeking to move its worst securities off its books and into the hands of unsuspecting clients. Without question, Merrill Lynch was doing that, especially with the triple-Bs. In one of the seamier examples of Merrill’s efforts to unload some of the junk on its balance sheet, it actually securitized subprime loans from Ownit—Bill Dallas’s subprime originator, which it partially owned—after Ownit filed for bankruptcy. Then again, every other big CDO underwriter on Wall Street—Citibank, UBS, Morgan Stanley, you name it—was doing the exact same thing. “People on the outside thought the market was going gangbusters because of all the deals getting done,” CDO expert Gene Phillips told Bloomberg. “People on the inside knew it was a last-gasp effort to clear out the warehouses.”

  In the aftermath of the crisis, Goldman Sachs would be the firm that was by far the most criticized for selling its clients down the river in its efforts to get risk off its own books. In truth, Goldman was just better at it than Merrill and the others. It was tougher and smarter in the way it went about it. And there was an even bigger difference between the way Merrill and Goldman went about attempting to reduce risk. Goldman as an institution never believed that the tiny bit of extra return offered by triple-A subprime-backed securities was worth the risk. As it began marking down its securities—and pushing them off its books—it treated triple-As just as ruthlessly as it treated all the other subprime securities it was marking.

  On May 16, 2007, Dow Kim announced that he was leaving Merrill Lynch to start a hedge fund; finally O’Neal said he could go.18 During the previous three years, the firm’s trading revenues had doubled; in 2006, his last full year with the firm, Kim was Merrill’s second highest-paid executive, after only O’Neal, taking home a paycheck of $37 million. Along with Fakahany, O’Neal had always viewed Kim as part of his inner circle and was gracious about his departure. It was only after the crisis that O’Neal would reflect back on Kim’s sudden departure, wondering why his head of fixed income hadn’t seen the problem coming. Or, worse, O’Neal would think in his darkest moments, maybe he had seen it coming. Maybe that’s why Kim had left.

  This seems unlikely. Semerci would later insist that he had shown Kim his risk positions, according to several former executives. But people who have seen the e-mail traffic say that that doesn’t appear to be the case. One day, several months after he had left the firm, Kim returned to Merrill’s headquarters, trying to rustle up a Merrill Lynch investment for his hedge fund. He ran into John Breit in the hallway. “It’s a debacle,” Breit told him, relating the enormous subprime exposure. Kim was stunned. “We don’t have all that stuff!” he replied. Truly, he hadn’t known.

  For that New Yorker article, O’Neal’s predecessor, David Komansky, told the writer John Cassidy that he simply didn’t believe O’Neal was unaware of the firm’s CDO exposure. Hard though it may be to believe, that does appear to be the case. “Stan was no longer dug in,” says a former executive. At the same time Goldman executives were canceling vacations to deal with the burgeoning subprime crisis, O’Neal was often on the golf course, playing round after round by himself. He had little or no direct contact with any of the firm’s operations—he had delegated that to Fleming, Fakahany, and others. Always a loner, he had become isolated from his own firm. He had no idea that key risk managers had been pushed aside, or that the people he had put in important positions were out of their depths. Amazing as it sounds, the CEO of Merrill Lynch really didn’t have a clue.

  In August, O’Neal went to Martha’s Vineyard for vacation. By then, the market was signaling that the end was near; on the ABX, even the triple As were starting to drop in value. In late July, the Dow had its worst week in more than four years. The CDO market continued to contract. Day after day, the decline continued. Somehow, the combination of the ongoing turmoil in the market and his ability to step back and see things more clearly while he was far away from Wall Street had the effect of finally rousing O’Neal. By the time he returned to work at the beginning of September, he was no longer in denial. O’Neal finally understood that the triple-A securities on Merrill’s book posed a huge threat to the firm. At a minimum, the securities were going to have to be marked down, and there would have to be write-downs that would damage Merrill’s earnings. The firm’s third-quarter earnings report was due in October; he had a month to come to terms with the problem. As he thought about it, O’Neal wasn’t just worried. The memory of the LTCM disaster was flooding back. He was scared.

  John Breit understood the problem by then as well. In July, Lattanzio had commandeered two junior quants and told them to sign off on a new valuation method the mortgage desk wanted to use for CDOs squared. The quants, feeling they were being asked to ratify something that had not been vetted through proper channels, complained to their manager. The manager happened to tell Breit the story. Breit’s curiosity was sparked. Calling in a favor from someone in the finance department, he got ahold of a spreadsheet with the collateral in the CDOs squared. He quickly saw how bad it was. He keep digging, quietly; before long he had discovered the $55 billion exposure.

  But Breit was still persona non grata on the trading floor. He had no access to top management. He had long since been tossed off the risk management committee. Thus he resorted to the only action he could think to take: he began buttonholing people he bumped into at Merrill, telling them the losses on the mortgage desk were going to be in the billions, not the millions. In early August, Breit went on vacation in the Hamptons. One day he received a phone call from Semerci,
who had heard through the grapevine what Breit was saying. Semerci was enraged, and insisted that the losses were only going to amount to a couple of hundred million dollars. By the end of August, the mortgage desk had upped its loss esimate to $600 million—a number Breit still thought was absurdly low.

  By mid-September, Semerci and Lattanzio were conceded $1.3 billion in triple-A losses. Seeing the problems grow, Greg Fleming reached out to his old friend Jeff Kronthal. O’Neal had named Fleming co-president of Merrill Lynch—along with Fakahany—shortly after Dow Kim left. Although he was still under strict orders to stay away from fixed income, the problems on the mortgage desk seemed too deep to just look the other way. Kronthal explained to Fleming how CDOs work and began tapping into his own sources at Merrill Lynch to see if he could find out what was going on. One of those sources was Breit. Breit told Kronthal that he thought the write-downs were going to be much bigger than anyone on the mortgage desk was admitting, which by then was around $3 billion. Kronthal conveyed this to Fleming, who conveyed it to O’Neal. O’Neal asked to see Breit.

  The two men had known each other for more than a dozen years; they had even worked together on occasion. O’Neal knew that Breit understood risk as well as anyone at Merrill. “I hear you have a model of the CDOs that disagrees with the valuations being put out there by Semerci,” O’Neal began. No, Breit replied, he didn’t have a model; just a back-of-the-envelope calculation. Then he gave O’Neal his number: $6 billion in losses. And he added, “It could be a lot worse. I haven’t even looked at the high-grade CDOs, just the CDOs squared and the mezzanines.”

  O’Neal looked like he was going to throw up. “What about all the protection we bought?” he asked. Breit explained that with AIG no longer in the business, Merrill had been buying protection from the monolines, which had taken on so much risk they would be insolvent long before they could pay off Merrill. O’Neal kept probing. What about the risk models? he asked. Worthless, replied Breit matter-of-factly. The risk wasn’t captured by VaR, and the VaR analysis of the underlying credit quality was wrong. Other risk models didn’t do any better. As O’Neal listened in silence, Breit explained how an important Merrill risk measure had been changed in such a way as to disguise the increasing amount of triple-A risk on the firm’s books. Breit today says he does not believe this was purposely changed to hide the ball—he thinks it might have even been a regulatory change—but it had that effect. “It distorted the true nature of the risk,” he told O’Neal. After talking for a few more minutes, Breit shook O’Neal’s hand and wished him luck. “I hope we talk again,” he said.