A flawed structure

Mortgage lenders active in the US subprime market have suffered a major reversal of fortune. Once considered the hottest sector of the mortgage market, sharply rising default rates in recent months have caused some of the largest names in the subprime area to write off losses and even file for bankruptcy. Just under 30 mortgage subprime lenders have been forced to close shop since January 2006 - and more are likely to follow.

New Century Financial, a California-based real estate investment trust and a major subprime mortgage lender, is one of the latest firms to reveal significant losses. In a February 7 conference call, the firm warned that its loan production would decrease by 20%. The mortgage originator blamed a trend of early defaults that started in the second quarter of 2006 and intensified later in the year. According to LoanPerformance, a San Francisco-based research and analytics firm, foreclosures - considered to be a conservative estimate of defaults - jumped from 2.22% in December 2005 to 3.33% a year later.

Further adding to its woes, New Century Financial disclosed in March that the US Attorney's Office for the Central District of California is undertaking a criminal probe of its trading activities and possible accounting errors. It is also in default with a number of its lines of credit, and its buyback provisions have been triggered. According to a March 19 regulatory filing, the company has $8.2 billion in repurchase obligations related to its credit facility agreements with other banks, including Barclays Capital and Morgan Stanley. Its share price subsequently plummeted from a closing price of $5.16 on March 7 to $0.85 on March 13, prompting the New York Stock Exchange to immediately suspend trading of the stock.

HSBC has also encountered difficulties in the subprime market. On February 22, chief executive Michael Geoghegan ousted Bobby Mehta, chief executive of HSBC North America Holdings and HSBC Finance Corporation, the bank's residential mortgage lending arm, following an announcement that the firm would have to write off about $10.56 billion in loan losses in its upcoming 2006 annual report - $1.8 billion more than analysts initially thought. In essence, the bank wrongly predicted how rising interest rates and falling housing prices would affect borrowers' payments.

The rising losses have alarmed many analysts who fear there could be a possible contagion effect from the subprime market in the overall housing market. So great is the worry that national regulators have stepped in. "The rapid rise in the default risk has caused concern. Whenever we see a 45-degree angle (upward trajectory in the number of subprime defaults), it causes supervisory concern," says a Washington, DC-based regulator.

On March 2, five regulators - the US Federal Reserve, the Federal Deposit Insurance Corporation, the National Credit Union Administration, the Office of the Comptroller of the Currency and the Office of Thrift Supervision - told US mortgage lenders they need to address certain risks relating to subprime mortgage lending practices. In particular, lenders will need to analyse borrowers' ability to repay debt by final maturity at a fully indexed rate, assuming a fully amortising repayment schedule, as opposed to basing their decisions on borrowers' ability to pay a low initial teaser rate (see box).

So how have mortgage lenders determined who to lend to, and why have so many underestimated potential losses? For a start, discrepancies in the inputs of models that predict default rates have led to big differences in credit risk forecasts, and this has translated into large losses for some banks. Inputs include combined loan-to-value ratios, current and future rates of interest, assumptions on house price appreciation, information on the type of house being purchased (single family, multi-family, condo and so on), and a borrower's Fico credit score. Fico scores, created by the Minneapolis-based Fair Isaac Corporation, rank consumers on the likelihood that they will make payments as expected. So, a borrower's Fico score represents his or her relative creditworthiness compared with the total population of borrowers. Fico scores range between 300 and 850.
However, analysts say the standard variables they relied upon to understand default risk in loan portfolios started to lose their predictive power in early 2006. "People are concerned that fundamental credit variables that people used to have faith in to have maximum predictability are unable to capture the risks of a loan," says Tanmoy Kumar Mukherjee, a New York-based senior vice-president at Sorin Capital Management, a $622 million real estate-focused hedge fund based in New York.

According to data from LoanPerformance, average debt-to-income levels for those borrowers taking out subprime adjustable-rate mortgages rose from 38.9 in the third quarter of 2000 to 42.1 in the third quarter of 2006. However, lenders thought the increased level of debt among subprime borrowers was being offset by the growing creditworthiness of borrowers. Average Fico scores rose from 594 to 625 over the same six-year period.

Rather than cancel each other out, default rates have risen sharply. Delinquencies for subprime mortgages reached 14.57% as of December 2006. That compares with 9.88% six years earlier, according to LoanPerformance (see figure 1).
An over-reliance on consumer credit scores is one of the main reasons some mortgage banks underestimated the level of delinquencies, say analysts. According to Dale Westhoff, New York-based head of mortgage-backed securities research at Bear Stearns, recent research on collateral risk by his bank shows that Fico scores are now not as predictive of default as had previously been thought.

"We found a Fico score has become less predictive of future default than documented income levels," said Westhoff, speaking at a Bear Stearns conference on subprime mortgages on March 9 in New York. Bear Stearns is now basing more of its credit risk analysis on documented income, rather than on Fico scores.

Skewed
Sorin Capital has also decreased the weighting of Fico scores in its credit risk models. Mukherjee says Fico data has been skewed in recent years by strong home price appreciation and low interest rates. Since 2000, borrowers with low Fico scores have been able to take out second or third loans at low rates of interest, or have been able to refinance existing loans because the value of their property has increased - therefore avoiding default. In 2000, a Fico score of 580 for a subprime mortgage borrower would imply cumulative losses of around 5-6%, holding all else equal, says Mukherjee. By 2006, that same credit score implied cumulative losses of only 3-4%.

There are also many unknown factors that have led banks to underestimate loan losses. "The market has no precedent. You don't have any history on this kind of market. It is so different that you can't base your default probabilities on your historic models," says Mukherjee.

For instance, house prices have experienced an unprecedented period of sustained growth over the past six years. From 2000-2006, housing prices jumped by 55.21%, according to a December 2006 quarterly housing report by the US Office of Federal Housing Enterprise Oversight. Since 1980, prices have risen by 307.51%.

Mark Liu, a Missouri-based director in the consumer analytics and modelling unit at Citi, says the greatest challenge has been predicting when and at what speed housing price appreciation would slow. With appreciating home prices, borrowers have been able to secure larger and larger loans based on the increased value of their houses. "Before, you may have input a house price appreciation of 6%. Now, you have to change that input to 1% or even -2% in some markets," he explains.
Speaking at the March 9 conference, Gyan Sinha, head of research for asset-backed securities at Bear Stearns in New York, said lenders were partially to blame for the increase in default risk. In particular, increased competition among mortgage originators and the need to ensure a steady flow of loans played a key part in rising defaults. "Last year and 2005 have not generally created a good vintage. The reason is that this is an industry that is notoriously prone to overcapacity. The barriers to entry are relatively low," said Sinha. "Was the industry set up so that someone could go into the housing market with a fundamentally bad business model that created perverse incentives to make bad loans?"

A major consequence of such fierce competition has been a reduction in underwriting standards. Some analysts have argued that flexible underwriting standards are not necessarily a bad thing, especially if more people end up owning homes (see box). But in terms of understanding how they affect default risk, reduced underwriting standards are generally viewed negatively.
Stated income loans, in which a borrower does not have to provide proof of his or her income, are highlighted by many analysts as particularly problematic. Piggyback loans or so-called 'silent seconds' provide another good example. The product - where a borrower receives a loan with a loan-to-value (LTV) ratio of 80% and simultaneously receives another loan with an LTV of 20% - allows a borrower to enter into a mortgage without putting down any of his or her own money. It means LTV ratios, a common input in default models, can no longer be relied upon.

"People generally focused on the LTV and were a little less focused on the behavioural impact of having a second loan, even if the borrower had less of an equity stake in the property," explains one managing director in charge of risk management on a mortgage trading desk at a leading Wall Street firm. "But firms have been originating more silent seconds than they did in the past, and as a result, combined loan-to-value ratios have got a lot more focus."
However, Mark Beardsell, a director in the modelling division at LoanPerformance, argues that the predictive power of combined loan-to-value ratios only goes so far. Many firms are not able to connect a borrower's predicted behaviour on his or her first loan with the details, pricing and risk management of the second loan because of internal divisions within banks and the limitations of their computer networks. Accordingly, many banks, including Countrywide and Bear Stearns, have dramatically reduced originating piggyback and no-money-down loans.
An even greater disconnect inside troubled banks is between their risk management and origination divisions. The Washington, DC-based regulator says a link between credit risk modelling and new loan origination does not exist at some lenders. "If they stress tested a realistic exercise and saw that losses would be 20% of their portfolio, then they should tie those results to what they do in their underwriting and to their reserves and what they planned for the business. But for some banks, these tests were just a theoretical exercise," he says.
All banks who spoke to Risk declined to comment on how their prepayment and default predictions informed their underwriting standards. Some banks, however, started pulling out of the subprime market before default rates started skyrocketing. For example, Citi decided to stop offering teaser rate and interest-only mortgages to low Fico score customers two years ago, according to its 2006 annual report. The decision resulted in a 20% decrease in its non-prime mortgage originations from 2004 to 2005.

But why does this disconnect exist in the first place? Beardsell believes that profit motivations tend to win out over sound risk management during periods of irrational exuberance. The argument goes as follows: if most mortgage originators have similar underwriting standards and one lender, regardless of its reasons, decides not to originate a certain product, then a competitor will quickly come in and snatch market share. Beardsell likens the situation to a party that ends with a game of musical chairs: "When the music stops, the question becomes: who can put on the brakes first?"

BOX
GIVING CREDIT WHERE NO CREDIT IS DUE
On March 2, US regulators requested public comment on a proposal regarding tightening-up underwriting standards in the subprime adjustable rate mortgage (Arm) lending business.

One Washington, DC-based regulator says many banks used the lowest possible standards when determining eligibility of a borrower. "Some lenders didn't fully analyse the borrower's capability to repay the loan based on the complete terms of the loan. They used only the initial teaser rate. That was prevalent in many institutions," he says. "If you know that the rates will jump up after a certain amount of time, you should make sure that the borrower can pay back the loans under the worst-case scenario."

According to the proposal, a financial institution must analyse a borrower's ability to repay the debt by its final maturity at a fully indexed rate, assuming a fully amortised repayment scheduled. The institution must also assess the borrower's total debt-to-income (DTI) ratio, which should equal the total monthly housing-related payments - including principal, interest, taxes and insurance - as a percentage of gross monthly income, and not just interest payments as a percentage of income. Furthermore, the regulator said a 60-65% DTI under normal circumstances would be a potential red flag.

The Washington, DC regulator cites sloppy risk management as another concern. Some banks are not scenario testing at all, while others are scenario-testing but don't link the results back to their loan portfolio, he says: "Stress-test results should go back to how they manage their business."

The regulators involved - US Federal Reserve, the Federal Deposit Insurance Corporation, the National Credit Union Administration, the Office of the Comptroller of the Currency and the Office of Thrift Supervision - are debating whether more regulation is necessary in the subprime mortgage sector. However, the outcome is by no means certain.

Speaking to Risk, John Dugan, comptroller of the Office of the Comptroller of the Currency, emphasises the need for all types of customers to have access to the markets. "We think additional guidance is necessary to address abuses in this market, but we also have to be careful not to impose a regulatory standard that goes too far and ends up denying credit to creditworthy borrowers - or preventing them from refinancing," says Dugan. "It is crucial to have public comment to give us more feedback on the potential impact of the proposed guidance."
Comments are due by May 7.

Risk