Tuesday, April 22, 2008

Deep inside the subprime crisis

Moody's walks Roger Lowenstein (writing for the Times Sunday magazine) through the construction, rating and demise of a pool of subprime mortgage securities. Some readers may have thought the IMF was exaggerating when it forecast up to $1 trillion in future losses from the credit bubble. After reading the following you will see that it's not an implausible number, and it will be clear why the system is paralyzed in dealing with (marking to market) the complicated securities (CDOs, etc.) that are contaminating the balance sheets of banks, investment banks, hedge funds, pension funds, sovereign wealth funds, etc. around the world.

Here's a quick physicist's calculation: roughly 10 million houses sold per year, assume that 10% of these mortgages are bad and will cost the issuer $100k to foreclose and settle. That means $100B per year in losses. Over the whole bubble, perhaps $300-500B in losses, which is more or less what the IMF estimates as the residential component of credit bubble losses (the rest of the trillion comes from commercial and corporate lending and consumer credit).

The internet bubble, with irrational investors buying shares of pet food e-commerce companies, was crazy. Read the excerpts below and you'll see that our recent housing boom was even crazier and at an unimaginably larger scale. (Note similar bubbles in the UK, Spain and in China.)

The best predictor, going forward, of mortgage default rates (not just subprime, but even prime mortgages) in a particular region will likely be the decline in home prices in that region. The incentive for a borrower to default on his or her mortgage is the amount by which they are "upside down" on the loan -- the amount by which their indebtedness exceeds the value of the home. Since we can't forecast price declines very well -- indeed, it's a nonlinear problem, with more defaults leading to more price declines, leading to more defaults -- we can't price the derivative securities built from those mortgages.

Efficient markets! ;-)

The figure above compares Case-Shiller data on the current bust (magenta) to the bust of the 80s-90s (blue). (Click for larger version.) You can see we have some way to go before all the fun ends.

Wall Street (Oliver Stone):

Gekko: Greed, for lack of a better word, is good. Greed is right. Greed works. Greed clarifies and cuts through and captures the essence of evolutionary spirit. Greed in all of its forms, greed for life, for money, for love, knowledge has marked the upward surge of mankind. And greed, you mark my words, will not only save Teldar Paper, but that other malfunctioning corporation called the USA.

Gekko: The richest one percent of this country owns half our country's wealth, five trillion dollars. One third of that comes from hard work, two thirds comes from inheritance, interest on interest accumulating to widows and idiot sons and what I do, stock and real estate speculation. It's bullshit. You got ninety percent of the American public out there with little or no net worth. I create nothing. I own.

Carl Fox: Stop going for the easy buck and start producing something with your life. Create, instead of living off the buying and selling of others.

NYTimes: ...The business of assigning a rating to a mortgage security is a complicated affair, and Moody’s recently was willing to walk me through an actual mortgage-backed security step by step. I was led down a carpeted hallway to a well-appointed conference room to meet with three specialists in mortgage-backed paper. Moody’s was fair-minded in choosing an example; the case they showed me, which they masked with the name “Subprime XYZ,” was a pool of 2,393 mortgages with a total face value of $430 million.

Subprime XYZ typified the exuberance of the age. All the mortgages in the pool were subprime — that is, they had been extended to borrowers with checkered credit histories. In an earlier era, such people would have been restricted from borrowing more than 75 percent or so of the value of their homes, but during the great bubble, no such limits applied.

Moody’s did not have access to the individual loan files, much less did it communicate with the borrowers or try to verify the information they provided in their loan applications. “We aren’t loan officers,” Claire Robinson, a 20-year veteran who is in charge of asset-backed finance for Moody’s, told me. “Our expertise is as statisticians on an aggregate basis. We want to know, of 1,000 individuals, based on historical performance, what percent will pay their loans?”

The loans in Subprime XYZ were issued in early spring 2006 — what would turn out to be the peak of the boom. They were originated by a West Coast company that Moody’s identified as a “nonbank lender.” Traditionally, people have gotten their mortgages from banks, but in recent years, new types of lenders peddling sexier products grabbed an increasing share of the market. This particular lender took the loans it made to a New York investment bank; the bank designed an investment vehicle and brought the package to Moody’s.

Moody’s assigned an analyst to evaluate the package, subject to review by a committee. The investment bank provided an enormous spreadsheet chock with data on the borrowers’ credit histories and much else that might, at very least, have given Moody’s pause. Three-quarters of the borrowers had adjustable-rate mortgages, or ARMs — “teaser” loans on which the interest rate could be raised in short order. Since subprime borrowers cannot afford higher rates, they would need to refinance soon. This is a classic sign of a bubble — lending on the belief, or the hope, that new money will bail out the old.

Moody’s learned that almost half of these borrowers — 43 percent — did not provide written verification of their incomes. The data also showed that 12 percent of the mortgages were for properties in Southern California, including a half-percent in a single ZIP code, in Riverside. That suggested a risky degree of concentration.

On the plus side, Moody’s noted, 94 percent of those borrowers with adjustable-rate loans said their mortgages were for primary residences. “That was a comfort feeling,” Robinson said. Historically, people have been slow to abandon their primary homes. When you get into a crunch, she added, “You’ll give up your ski chalet first.”

Another factor giving Moody’s comfort was that all of the ARM loans in the pool were first mortgages (as distinct from, say, home-equity loans). Nearly half of the borrowers, however, took out a simultaneous second loan. Most often, their two loans added up to all of their property’s presumed resale value, which meant the borrowers had not a cent of equity.

In the frenetic, deal-happy climate of 2006, the Moody’s analyst had only a single day to process the credit data from the bank. The analyst wasn’t evaluating the mortgages but, rather, the bonds issued by the investment vehicle created to house them. A so-called special-purpose vehicle — a ghost corporation with no people or furniture and no assets either until the deal was struck — would purchase the mortgages. Thereafter, monthly payments from the homeowners would go to the S.P.V. The S.P.V. would finance itself by selling bonds. The question for Moody’s was whether the inflow of mortgage checks would cover the outgoing payments to bondholders. From the investment bank’s point of view, the key to the deal was obtaining a triple-A rating — without which the deal wouldn’t be profitable. That a vehicle backed by subprime mortgages could borrow at triple-A rates seems like a trick of finance. “People say, ‘How can you create triple-A out of B-rated paper?’ ” notes Arturo Cifuentes, a former Moody’s credit analyst who now designs credit instruments. It may seem like a scam, but it’s not.

The secret sauce is that the S.P.V. would float 12 classes of bonds, from triple-A to a lowly Ba1. The highest-rated bonds would have first priority on the cash received from mortgage holders until they were fully paid, then the next tier of bonds, then the next and so on. The bonds at the bottom of the pile got the highest interest rate, but if homeowners defaulted, they would absorb the first losses.

It was this segregation of payments that protected the bonds at the top of the structure and enabled Moody’s to classify them as triple-A. Imagine a seaside condo beset by flooding: just as the penthouse will not get wet until the lower floors are thoroughly soaked, so the triple-A bonds would not lose a dime unless the lower credits were wiped out.

Structured finance, of which this deal is typical, is both clever and useful; in the housing industry it has greatly expanded the pool of credit. But in extreme conditions, it can fail. The old-fashioned corner banker used his instincts, as well as his pencil, to apportion credit; modern finance is formulaic. However elegant its models, forecasting the behavior of 2,393 mortgage holders is an uncertain business. “Everyone assumed the credit agencies knew what they were doing,” says Joseph Mason, a credit expert at Drexel University. “A structural engineer can predict what load a steel support will bear; in financial engineering we can’t predict as well.”

Mortgage-backed securities like those in Subprime XYZ were not the terminus of the great mortgage machine. They were, in fact, building blocks for even more esoteric vehicles known as collateralized debt obligations, or C.D.O.’s. C.D.O.’s were financed with similar ladders of bonds, from triple-A on down, and the credit-rating agencies’ role was just as central. The difference is that XYZ was a first-order derivative — its assets included real mortgages owned by actual homeowners. C.D.O.’s were a step removed — instead of buying mortgages, they bought bonds that were backed by mortgages, like the bonds issued by Subprime XYZ. (It is painful to consider, but there were also third-order instruments, known as C.D.O.’s squared, which bought bonds issued by other C.D.O.’s.)

Miscalculations that were damaging at the level of Subprime XYZ were devastating at the C.D.O. level. Just as bad weather will cause more serious delays to travelers with multiple flights, so, if the underlying mortgage bonds were misrated, the trouble was compounded in the case of the C.D.O.’s that purchased them.

Moody’s used statistical models to assess C.D.O.’s; it relied on historical patterns of default. This assumed that the past would remain relevant in an era in which the mortgage industry was morphing into a wildly speculative business. The complexity of C.D.O.’s undermined the process as well. Jamie Dimon, the chief executive of JPMorgan Chase, which recently scooped up the mortally wounded Bear Stearns, says, “There was a large failure of common sense” by rating agencies and also by banks like his. “Very complex securities shouldn’t have been rated as if they were easy-to-value bonds.”

...The challenge to investment banks is to design securities that just meet the rating agencies’ tests. Risky mortgages serve their purpose; since the interest rate on them is higher, more money comes into the pool and is available for paying bond interest. But if the mortgages are too risky, Moody’s will object. Banks are adroit at working the system, and pools like Subprime XYZ are intentionally designed to include a layer of Baa bonds, or those just over the border. “Every agency has a model available to bankers that allows them to run the numbers until they get something they like and send it in for a rating,” a former Moody’s expert in securitization says. In other words, banks were gaming the system; according to Chris Flanagan, the subprime analyst at JPMorgan, “Gaming is the whole thing.”

When a bank proposes a rating structure on a pool of debt, the rating agency will insist on a cushion of extra capital, known as an “enhancement.” The bank inevitably lobbies for a thin cushion (the thinner the capitalization, the fatter the bank’s profits). It’s up to the agency to make sure that the cushion is big enough to safeguard the bonds. The process involves extended consultations between the agency and its client. In short, obtaining a rating is a collaborative process.

The evidence on whether rating agencies bend to the bankers’ will is mixed. The agencies do not deny that a conflict exists, but they assert that they are keen to the dangers and minimize them. For instance, they do not reward analysts on the basis of whether they approve deals. No smoking gun, no conspiratorial e-mail message, has surfaced to suggest that they are lying. But in structured finance, the agencies face pressures that did not exist when John Moody was rating railroads. On the traditional side of the business, Moody’s has thousands of clients (virtually every corporation and municipality that sells bonds). No one of them has much clout. But in structured finance, a handful of banks return again and again, paying much bigger fees. A deal the size of XYZ can bring Moody’s $200,000 and more for complicated deals. And the banks pay only if Moody’s delivers the desired rating. Tom McGuire, the Jesuit theologian who ran Moody’s through the mid-’90s, says this arrangement is unhealthy. If Moody’s and a client bank don’t see eye to eye, the bank can either tweak the numbers or try its luck with a competitor like S.&P., a process known as “ratings shopping.”

...From 2002 to 2006, Moody’s profits nearly tripled, mostly thanks to the high margins the agencies charged in structured finance. In 2006, Moody’s reported net income of $750 million. Raymond W. McDaniel Jr., its chief executive, gloated in the annual report for that year, “I firmly believe that Moody’s business stands on the ‘right side of history’ in terms of the alignment of our role and function with advancements in global capital markets.”

...Moody’s was aware that mortgage standards had been deteriorating, and it had been demanding more of a cushion in such pools. Nonetheless, its credit-rating model continued to envision rising home values. Largely for that reason, the analyst forecast losses for XYZ at only 4.9 percent of the underlying mortgage pool. Since even the lowest-rated bonds in XYZ would be covered up to a loss level of 7.25 percent, the bonds seemed safe.

XYZ now became the responsibility of a Moody’s team that monitors securities and changes the ratings if need be (the analyst moved on to rate a new deal). Almost immediately, the team noticed a problem. Usually, people who finance a home stay current on their payments for at least a while. But a sliver of folks in XYZ fell behind within 90 days of signing their papers. After six months, an alarming 6 percent of the mortgages were seriously delinquent. (Historically, it is rare for more than 1 percent of mortgages at that stage to be delinquent.)

Moody’s monitors began to make inquiries with the lender and were shocked by what they heard. Some properties lacked sod or landscaping, and keys remained in the mailbox; the buyers had never moved in. The implication was that people had bought homes on spec: as the housing market turned, the buyers walked.

By the spring of 2007, 13 percent of Subprime XYZ was delinquent — and it was worsening by the month. XYZ was hardly atypical; the entire class of 2006 was performing terribly. (The class of 2007 would turn out to be even worse.)

In April 2007, Moody’s announced it was revising the model it used to evaluate subprime mortgages. It noted that the model “was first introduced in 2002. Since then, the mortgage market has evolved considerably.” This was a rather stunning admission; its model had been based on a world that no longer existed.

Poring over the data, Moody’s discovered that the size of people’s first mortgages was no longer a good predictor of whether they would default; rather, it was the size of their first and second loans — that is, their total debt — combined. This was rather intuitive; Moody’s simply hadn’t reckoned on it. Similarly, credit scores, long a mainstay of its analyses, had not proved to be a “strong predictor” of defaults this time. Translation: even people with good credit scores were defaulting. Amy Tobey, leader of the team that monitored XYZ, told me, “It seems there was a shift in mentality; people are treating homes as investment assets.” Indeed. And homeowners without equity were making what economists call a rational choice; they were abandoning properties rather than make payments on them. Homeowners’ equity had never been as high as believed because appraisals had been inflated.

Over the summer and fall of 2007, Moody’s and the other agencies repeatedly tightened their methodology for rating mortgage securities, but it was too late. They had to downgrade tens of billions of dollars of securities. By early this year, when I met with Moody’s, an astonishing 27 percent of the mortgage holders in Subprime XYZ were delinquent. Losses on the pool were now estimated at 14 percent to 16 percent — three times the original estimate. Seemingly high-quality bonds rated A3 by Moody’s had been downgraded five notches to Ba2, as had the other bonds in the pool aside from its triple-A’s.

The pain didn’t stop there. Many of the lower-rated bonds issued by XYZ, and by mortgage pools like it, were purchased by C.D.O.’s, the second-order mortgage vehicles, which were eager to buy lower-rated mortgage paper because it paid a higher yield. As the agencies endowed C.D.O. securities with triple-A ratings, demand for them was red hot. Much of it was from global investors who knew nothing about the U.S. mortgage market. In 2006 and 2007, the banks created more than $200 billion of C.D.O.’s backed by lower-rated mortgage paper. Moody’s assigned a different team to rate C.D.O.’s. This team knew far less about the underlying mortgages than did the committee that evaluated Subprime XYZ. In fact, Moody’s rated C.D.O.’s without knowing which bonds the pool would buy.

A C.D.O. operates like a mutual fund; it can buy or sell mortgage bonds and frequently does so. Thus, the agencies rate pools with assets that are perpetually shifting. They base their ratings on an extensive set of guidelines or covenants that limit the C.D.O. manager’s discretion.

Late in 2006, Moody’s rated a C.D.O. with $750 million worth of securities. The covenants, which act as a template, restricted the C.D.O. to, at most, an 80 percent exposure to subprime assets, and many other such conditions. “We’re structure experts,” Yuri Yoshizawa, the head of Moody’s’ derivative group, explained. “We’re not underlying-asset experts.” They were checking the math, not the mortgages. But no C.D.O. can be better than its collateral.

Moody’s rated three-quarters of this C.D.O.’s bonds triple-A. The ratings were derived using a mathematical construct known as a Monte Carlo simulation — as if each of the underlying bonds would perform like cards drawn at random from a deck of mortgage bonds in the past. There were two problems with this approach. First, the bonds weren’t like those in the past; the mortgage market had changed. As Mark Adelson, a former managing director in Moody’s structured-finance division, remarks, it was “like observing 100 years of weather in Antarctica to forecast the weather in Hawaii.” And second, the bonds weren’t random. Moody’s had underestimated the extent to which underwriting standards had weakened everywhere. When one mortgage bond failed, the odds were that others would, too.

Moody’s estimated that this C.D.O. could potentially incur losses of 2 percent. It has since revised its estimate to 27 percent. The bonds it rated have been decimated, their market value having plunged by half or more. A triple-A layer of bonds has been downgraded 16 notches, all the way to B. Hundreds of C.D.O.’s have suffered similar fates (most of Wall Street’s losses have been on C.D.O.’s). For Moody’s and the other rating agencies, it has been an extraordinary rout.

...The agencies have blamed the large incidence of fraud, but then they could have demanded verification of the mortgage data or refused to rate securities where the data were not provided. That was, after all, their mandate. This is what they pledge for the future. Moody’s, S.&P. and Fitch say that they are tightening procedures — they will demand more data and more verification and will subject their analysts to more outside checks. None of this, however, will remove the conflict of interest in the issuer-pays model. Though some have proposed requiring that agencies with official recognition charge investors, rather than issuers, a more practical reform may be for the government to stop certifying agencies altogether. ...


Banks, Bill said...

Polluting the balance sheets of banks, investment banks, hedge funds, pension funds, sovereign wealth funds, etc. around the world is not good. It is the bank who helps the people and also solves there financial problems. Banks should be on the save side.

Anonymous said...

So who is "Subprime XYZ"? Anyone want to speculate?

Roger Bigod said...

I couldn't tell from Lowenstein's piece what kind of mathematical procedure Moody's was using to evaluate "XYZ". Best I could tell, there were no path integrals, assumptions about tails, etc. It looked like they applied historical data with a little pro forma Monte Carlo simulation. The intellectual roots were in the green-eyeshades world of the 19th Cent. insurance industry, not diffusion equations. They just applied solid historical data in a situation where circumstances had changed.

If the history of this is ever written, I suspect the sophisticated models will have mixed results. Applied blindly, they were associated with some funds that blew up. OTOH, Simons was burned in August but adapted to the situation quickly. As in physics, insight and intuition into the applicability of a model can really help out a lot. :-)

Blog Archive