The Quants: How a New Breed of Math Whizzes Conquered Wall Street and Nearly Destroyed It (32 page)

Brown, meanwhile, went back to full-time work at Citigroup in 2000, working on a firmwide risk management system for the largest bank in the world. He found that Citi had much of its risk under control. But one corner of the bank bothered him: securitization. Since the bank’s securitization activities took place “off balance sheet,” in offshore accounts, there was a disturbing lack of transparency. It was hard to know exactly what was going on, how much risk was being taken. There seemed little he could do about it, aside from complaining to management from time to time. But who would listen? The business was a profit juggernaut. Naysayers were ignored.

Brown watched as the relatively sedate financial system he’d joined in the 1980s turned into a derivatives-laden, debt-grinding monster. Banks were dabbling in the most exotic derivatives imaginable. Blowups were becoming more frequent, but they seemed dwarfed by the massive amounts of money coming in the door. The casino stayed open for business. Indeed, it started branching out, searching for more ways to pull in capital its traders could play around with. For instance, subprime mortgages.

Like most
Wall Streeters, however, Aaron Brown was dazzled by the numbers, by the ingenious trading strategies that could arb out inefficiencies and deliver seemingly endless profits. Indeed, virtually the entire quant community, aside from a few random party-poopers, embraced the derivatives explosion wholeheartedly. The layered levels of complexity didn’t bother them whatsoever. They loved it.

Perhaps the most egregious example of over-the-top quantitative creativity involved those synthetic CDOs such as J. P. Morgan’s Bistro. Because of the complexity of all of their tangled and tranched swaps and bonds, it was very difficult to price all of the pieces. The biggest problem was the one Weinstein focused on years later: correlation. If loans in one piece of the CDO weaken, what are the odds that loans
in other parts will see problems? It’s the same problem as asking whether all the apples in a bag will start to rot if a few go bad.

Naturally, a quant was waiting in the wings with an elegant solution to it all—a solution that would help drive global credit markets into a ditch several years later.

That solution came from a Chinese-born quant named David X. Li, a financial engineer at the New York headquarters of the Canadian Imperial Bank of Commerce, or CIBC. Rather than try to model all of the fiendishly difficult factors that make the pricing of all the interrelated pieces so thorny, Li hit upon a quick fix that would immediately provide the data to price the hodgepodge of CDO tranches.

Li often discussed the problem with colleagues from academia who were experts in an actuarial science called survival analysis. One concept they studied was the fact that after the death of a spouse, people tend to die sooner than their demographic peers. In other words, they were measuring correlations between the deaths of spouses.

The link between dying spouses and credit default swaps was quant wizardry at its best—and its worst. Li showed how this model could assign correlations between tranches of CDOs by measuring the price of credit default swaps linked to the underlying debt. Credit default swaps supply a single variable that incorporates the market’s assessment of how the loan will perform. The price of a CDS, after all, is simply a reflection of the view investors have on whether or not a borrower will default.

Li’s model supplied a method to bundle the prices of many different credit default swaps in a CDO and spit out numbers showing the correlations between the tranches. In April 2000, having moved on to J. P. Morgan’s credit department, he published his results in the
Journal of Fixed Income
in a paper called “On Default Correlation: A Copula Function Approach.” The model’s name was based in part on the statistical method he used to measure correlations: the Gaussian copula.

Copulas are mathematical functions that calculate the connections between two variables—in other words, how they “copulate.” When X happens (such as a homeowner defaulting), there’s a Y chance that
Z happens (a neighboring homeowner defaults). The specific copulas Li used were named after Carl Friedrich Gauss, the nineteenth-century German mathematician known for devising a method, based on the bell curve, to measure the motion of stars.

The correlations between the slices in a CDO were, therefore, based on the bell curve (a copula is essentially a multidimensional bell curve). Thousands of bonds (or the swaps linked to them) weren’t expected to make big, sudden jumps; rather, they were generally expected to move from one point to another, up or down, in relatively predictable patterns. Extreme moves in a large number of underlying bonds weren’t part of the model. It was the law of large numbers all over again, the same mathematical trick Ed Thorp used to beat blackjack in the 1960s and that Black and Scholes used to price options. Now, however, it was being applied on a scale so vast and complex that it approached the absurd. Undaunted, the quants lapped it up.

As the synthetic CDO market boomed, Wall Street and credit rating agencies adopted Li’s model. “The Gaussian copula was the Black-Scholes for credit derivates,” said Michel Crouhy, Li’s boss at CIBC in the 1990s. So-called correlation traders sprang up at banks such as Goldman Sachs, Morgan Stanley, and Deutsche Bank, using the model to trade CDO tranches, and the underlying correlations between them, like baseball cards. The model seemed to work relatively well and was easy to use.

Crucially, and disastrously, the model was based on how
other investors
viewed the market through the lens of credit default swaps. If CDS traders thought few homeowners would default on their loans, Li’s Gaussian copula priced the tranches accordingly. And since the CDO boom was occurring at the same time that a housing bubble was inflating—indeed, it helped inflate the bubble—most investors believed there was little chance that a large number of loans would default. What resulted was a vicious feedback loop—an echo chamber, one might say—in which enthusiastic investors snapped up tranches of CDOs, creating demand for more CDOs—and that created a demand for more mortgage loans. The CDOs were showing very little risk, according to Li’s model. For some reason, nearly everyone, except for a few doubting Thomases in the wilderness, believed in it,
even though the historical record of how mortgage loans behaved in a broad economic downturn was vanishingly slim.

Then, in 2004, to meet the insatiable demand, banks started packing CDOs with a type of loan Li hadn’t considered when creating his model in the late 1990s: subprime mortgages. The CDO market went into hyperdrive.

Thanks to even more quant alchemy, certain tranches of subprime CDOs could earn AAA ratings from agencies such as Standard & Poor’s, a stamp of approval that allowed regulated entities such as pension funds to gobble them up. Here’s how it worked: Financial engineers would take lower-rated slices of a mortgage-backed security or other securitized bundle of loans such as credit card lines, and package them in a CDO. It would then slice the CDO into different pieces, based on priority—which slices had the right to the cash spit out by the loans first, second, third, and so on. A product that began as home loans to the riskiest kind of borrower went through the looking glass of quantdom and came out a gold-plated security, suitable for some of the most closely watched and regulated investors. In fact, they were only low-risk relative to other, even more volatile tranches, when viewed through the rose-colored glasses of boom-time investors.

In 2004, $157 billion in CDOs was issued, much of which contained subprime mortgages. The amount spiked to $273 billion in 2005 and a whopping $550 billion in 2006, its peak year.

The Gaussian copula was, in hindsight, a disaster. The simplicity of the model hypnotized traders into thinking that it was a reflection of reality. In fact, the model was a jury-rigged formula based on the irrationally exuberant, self-reinforcing, and ultimately false wisdom of the crowd that assigned make-believe prices to an incredibly complex product. For a while it worked, and everyone was using it. But when the slightest bit of volatility hit in early 2007, the whole edifice fell apart. The prices didn’t make sense anymore. Since nearly every CDO manager and trader used the same formula to price the fizzing bundles—yet another instance of crowding that results from popular quant methodologies—they all imploded at once.

Is it any wonder why? The complexity had become malevolent. The quants and correlation traders were modeling cash flows on
tranches of credit default swaps tied to CDOs that were bundles of mortgage-backed securities, which in turn were tranched packages of opaque loans from homeowners around the country. The model created an illusion of order where none existed.

A key player in the CDO boom was a Citadel baby, a $5 billion hedge fund called Magnetar Capital run by one of Ken Griffin’s star traders, Alec Litowitz. In 2006,
Total Securitization
, an industry newsletter, named Magnetar investor of the year. “Magnetar bought bespoke deals in massive size in 2006, investing in a series of CDOs—each over $1 billion,” the newsletter said in March 2007.

Magnetar’s presence in the CDO world can be found in Litowitz’s seeming fascination with astronomy. A large number of toxic CDOs created at the height of the subprime frenzy had astronomical names, such as Orion, Aquarius, Scorpius, Carina, and Sagittarius. Magnetar was “their lynchpin investor,” according to an investigation by the
Wall Street Journal
. But Magnetar, which gained 25 percent in 2007, was also taking the other side of slices of those CDOs, buying positions that would pay off when higher-rated slices turned sour.

Magnetar’s trade was ingenious, and possibly diabolical. It would hold the riskiest slices of CDOs, known as the “equity”—those most vulnerable to defaults. But it also was buying protection on less-risky slices higher up the stack of the CDO’s structure, essentially betting on a wave of defaults. The roughly 20 percent yield on the equity slices provided the cash to purchase the less-risky slices. If the equity imploded, as it did, the losses would mean little if the higher-quality slices also saw significant losses, which they did.

In hindsight, Magnetar turned out to be a facilitator of the CDO boom, because it gobbled up those equity slices when few other investors wanted to buy them. Without a willing buyer of the junk slices, banks would have had a much harder time constructing the increasingly dicey CDOs hitting the market in 2006 and 2007. In all, Magnetar was a key investor in roughly $30 billion of constellation CDOs issued from mid-2006 to mid-2007.

There’s clear evidence that Wall Street’s gluttonous demand for loans and all the fat fees they spit out was the key factor that allowed, and encouraged, brokers to concoct increasingly risky mortgages with
toxic bells and whistles such as adjustable interest rates that shot higher a few years—or in some cases a few months—after the loan was made. Out of twenty-five of the top subprime mortgage lenders, twenty-one were either owned or financed by major Wall Street or European banks, according to a report by the Center for Public Integrity. Without the demand from the investment banks, the bad loans would never have been made.

As the CDO boom took off, so did home prices across the United States. From January 2000 through July 2006, the peak of the housing bubble, the average price of a home in the United States rose 106 percent, according to the S&P/Case-Shiller National Home Price Index. To models such as the Gaussian copula, the message was clear: the housing market was getting safer and safer. In fact, it was getting far more dangerous. In late 2006, the home price index started to move in the opposite direction, falling more than 30 percent three years later.

Some quants, including Brown himself, criticized the models that banks and credit rating agencies were using to price CDOs. He knew the correlations spat out by the Gaussian copula were a fantasy. But as long as the money was rolling in, no one wanted to hear it—not the correlation traders making fat bonuses, and definitely not the Wall Street CEOs making even fatter bonuses.

Like crack cocaine, it was addictive, and ultimately ruinous. While the boom lasted, securitization helped Wall Street become an increasingly powerful force in the U.S. economy. The financial sector’s share of total U.S. corporate profits hit 35 percent in 2007, up from 10 percent in the early 1980s, when quants such as Brown started to arrive on the scene. Financial institutions made up one-fourth of the market cap of the S&P 500, far more than any other industry.

Helping to drive the surge in financial profits was that clever tactic favored by funds such as AQR, Global Alpha, Citadel, and Saba: the carry trade. By late 2006, more money than ever had been plowed into the trade, in which investors, usually banks and hedge funds, borrowed low-yielding currencies such as Japanese yen to buy higher-yielding currencies such as the New Zealand dollar or British pound. It was a frictionless digital push-button cash machine based on math and computers—a veritable quant fantasyland of riches.

The carry trade was fueling a worldwide liquidity boom, sparking a frenzy in everything from commodities to real estate—and subprime mortgages. “They can borrow at near zero interest rates in Japan … to relend anywhere in the world that offers higher yields, whether Argentine notes or U.S. mortgage securities,” marveled the United Kingdom newspaper the
Telegraph
. “It has prolonged asset bubbles everywhere.”

“The carry trade has pervaded every single instrument imaginable, credit spreads, bond spreads: everything is poisoned,” HSBC currency analyst David Bloom told the paper.

Few, however, seemed to worry about what would happen if the trade suddenly fell apart. Every now and then, a hiccup foreshadowed the incredible turmoil to come. In February 2007, traders started getting nervous about whether stocks in China and other emerging markets had run up too far, too fast. As Chinese stocks started to fall, traders who’d plowed into the market using carry-trade rocket fuel started to panic, buying back their borrowed yen and causing the yen to spike.